Validation and Improvement of The ISO 2631-1 (1997) Standard Method For Evaluating Discomfort From Whole-Body Vibration in A Multi-Axis Environment

Loughborough University
Institutional Repository
Validation and improvement

of the ISO 2631-1 (1997)
standard method for
evaluating discomfort from
whole-body vibration in a
multi-axis environment
This item was submitted to Loughborough University's Institutional Repository
by the/an author.
Additional Information:
• This thesis is not yet available. A Doctoral Thesis. Submitted in partial

fulllment of the requirements for the award of Doctor of Philosophy of
Loughborough University.
Metadata Record: https://dspace.lboro.ac.uk/2134/6250
Publisher: c Yka Marjanen

Please cite the published version.

This item was submitted to Loughborough’s Institutional Repository
(https://dspace.lboro.ac.uk/) by the author and is made available under the
following Creative Commons Licence conditions.
For the full text of this licence, please go to:

http://creativecommons.org/licenses/by-nc-nd/2.5/
Validation and improvement of the ISO 2631-1
(1997) standard method for evaluating discomfort
from whole-body vibration in a multi-axis
environment
By
Ykä Marjanen
Submitted in partial fulfilment of the requirements for the award of

Doctor of Philosophy of Loughborough University
January 2010
© by Ykä Marjanen (2010)

Declaration
No portion of this work referred to in this thesis nor the original work contained therein
has been submitted to this or any other institution for a degree.
Signed .............................................
Ykä Marjanen
II
Loughborough University
Abstract
Doctor of Philosophy
Validation and improvement of the ISO 2631-1 (1997) standard method for
evaluating discomfort from whole-body vibration in a multi-axis
environment
Ykä Marjanen
Vibration exposure can occur at work, commuting between home and work, and in
leisure activities. Any form of transportation will expose humans to some degree
of vibration. Exposure to vibration can cause health problems, but more likely
comfort problems. Health problems are normally related to back pain. Comfort on the
other hand is related to both physiological and psychological factors, which can have a
wide range of effects from a general annoyance to a reduced work capability.
The standard ISO 2631-1 (1997) provides a guidance, which can be used to measure,
evaluate and assess effects of whole-body vibration to discomfort. The standard allows
several interpretations, which can lead to different results, as the standard does not
provide an explicit guidance for selecting which axes and locations to measure and
which averaging method to use for evaluating the axes. The suggested averaging
method is the root mean square (r.m.s.) method, but additionally vibration dose value
(VDV) can be used. This can lead to different results, as VDV emphasises shocks
more than the r.m.s. method. The standard guides to measure and evaluate at least
the seat translational axes, but the additional nine axes from the seat, backrest and
floor are not mandatory. However, this can result in a different comfort value, as the
values from the measured axes are combined. So taking into account all possible
interpretations the assessment can vary significantly for the same environment.
The selection of the averaging method is not a technical issue, as both methods are
supported by all commercial equipment. However, it is rare that more than three axes
are possible to be measured with typical whole-body vibration measurement
equipment, thus the majority of studies have published results based on only the seat
translational axes. Especially the rotational axes have been missing in most studies.
The full method (i.e. using all possible axes to calculate the comfort value) of ISO
2631-1 (1997) has been rarely used and there is very little information on how
accurate the method is for assessing discomfort in a multi-axis environment. There are
only a few studies that have used the full method, but there are no known studies
which have actually validated the full ISO 2631-1 method.
The objective of the thesis was to validate and, if necessary, to improve the full method
of the ISO 2631-1 standard for evaluating discomfort from whole-body vibration in a
multi-axis environment. It was assumed that the ISO 2631-1 method can be used to
predict discomfort in practice, but there are a relatively low number of studies to
confirm this. Frequency weightings have been the focus of many published studies and
it was assumed that these are broadly correct. Other aspects of the ISO 2631-1
method are the focus of this thesis. The goal was to keep a backward compatibility to
previous studies and the current commercial equipment, thus several limitations were
defined for the improvement of the standard.
III
Several laboratory experiments, field measurements, and field and laboratory trials
were conducted to validate the standard method. At first it was concluded that practical
equipment for measuring 12-axis data was needed as there was no commercial system
available. The equipment and software was validated in two experiments, which
showed that simple and affordable components could be used to develop equipment
for the full method. Even though the standard does not include information about a six-
axis sensor for measuring both translational and rotational axes, there was a method to
validate the sensor.
The first field study included measuring several machines using all twelve axes. The
analysis showed that the seat and backrest translational axes will contribute about 90
% of the overall vibration total value of the standard method, thus very little justification
was found for including the seat rotational and floor translational axes. Similar results
were found based on the data from the previous 12-axis studies. It was also found that
the neglected axes could be compensated with a factor for estimating the overall
vibration total value including all twelve axes. As the overall vibration total value is
directly related to the number of used axes, the compensating factors can be used to
compare results which used different axes.
The laboratory trial confirmed the results from the field study, and it was concluded that
sufficient accuracy to predict discomfort can be achieved using just the seat
translational axes, even though the correlation improved when more axes were
included. It was found that the evaluation of discomfort was improved by the use of the
frequency weighting curves and the r.m.s. averaging method. However, as the
multiplying factors degraded correlation, it was concluded that a new set of factors
should be calculated. The new factors showed that a higher emphasis on the seat
horizontal axes should be given (x=2.7, y=1.8 and z=1.0). The new factors improved
the correlation systematically for all subjects.
The field trial showed a similar trend, where optimised multiplying factors improved the
correlation, but it was also noted that different multiplying factors are required for
different environments, thus a procedure to optimise the standard method to different
environments was developed. The trial showed systematic behaviour and the
optimised multiplying factors were best for all subjects and groups.
Keywords: Discomfort, whole-body vibration, standard, ISO 2631-1, multi-axis,

multiplying factors
IV
Acknowledgements
At first I would like to thank my supervisor Neil Mansfield for giving me the support and
opportunity to work from abroad to Loughborough University. With your guidance I feel
like I have evolved to a true research scientist. I think our arrangement worked out
pretty well. Also the staff of Human Sciences Department and Environment
Ergonomics Laboratory have been supportive towards my efforts: especially George
Havenith as my Director of Research and Helen Relf for helping with the practical
arrangements. Also I would like to thank fellow PhD students and visitors at the
laboratory for helping me with practical arrangements and keeping me company:
Lauren, Sham and Robin.
It is a rare chance to meet someone like Dr Setsuo Maeda, who seems to have
endless energy. While doing so many other things he had time to exchange emails and
skype calls with me, arrange visitation to his institute (JNIOSH) and co-write
publications. I would also like to thank Dr Shibata Nobuyuki for practical arrangements
with the experiments in Japan. Your attitude towards work still amazes me.
I have been fortunate to have resources to participate in many conferences concerning

human response to vibration. Especially the UK conferences have been an excellent
venue. So thanks to all colleagues whom I have had the pleasure talking with.
It has been interesting four years. During the same period I have started a company,
worked on my thesis and started a family (two children and counting). My business
partner Janne Göös has supported me by helping in practical arrangements in the PhD
measurements and trials, and allowing me to spend time on working the thesis. Thank
you Janne!
My wife Ona, with whom I have shared over half of my life and created new lives, I
thank humbly. In a family, the work load never vanishes, but transfers between the
partners. You have done at least the equal amount of extra work to our family, that I
have done for the thesis. Thus you deserve the highest honor! Oskari and Noora,
respect your mother!
My parents and grand parents deserve also mentioning, as I am who I am because of

the genes and environment. So whatever you have done, has led me to this point.
I would like to thank The Finnish Work Environment Fund for supporting my PhD work
financially. Without the financial support this work would have taken much longer.
V
Table of Contents
ABSTRACT ....................................................................................................... III
ACKNOWLEDGEMENTS ................................................................................. V
TABLE OF CONTENTS ................................................................................... VI
LIST OF TABLES ........................................................................................... XVI
LIST OF FIGURES ......................................................................................... XXI
ABBREVIATIONS ......................................................................................XXVIII
1 CHAPTER 1 – INTRODUCTION .................................................................. 1
1.1 BACKGROUND .................................................................................................. 1

1.2 PROBLEM DESCRIPTION .................................................................................... 2
1.3 OBJECTIVES AND HYPOTHESES ......................................................................... 4
1.4 THESIS STRUCTURE .......................................................................................... 5
1.5 SCOPE............................................................................................................. 6
1.6 ORIGINAL CONTRIBUTION TO KNOWLEDGE .......................................................... 7
1.7 PUBLICATIONS ................................................................................................. 8
2 CHAPTER 2 - LITERATURE REVIEW ........................................................ 9
2.1 INTRODUCTION AND AIMS OF THE REVIEW .......................................................... 9

2.2 EFFECTS FROM VIBRATION EXPOSURE ............................................................. 10
2.2.1 Discomfort ........................................................................................................... 11
2.2.1.1 Studies and results behind the ISO 2631-1 (1997) method ............................ 14
2.2.1.2 Laboratory studies and results validating the standard method ..................... 22
2.2.1.3 Field studies and results validating the standard method ............................... 24
2.2.1.4 Issues relating to the understanding of discomfort effects from whole-body
vibration exposure ........................................................................................................... 27
2.2.2 Health .................................................................................................................. 28
2.2.2.1 Low back pain ................................................................................................. 29
2.2.2.2 Other effects .................................................................................................... 30
2.2.2.3 Issues relating to the understanding of health effects from whole-body
vibration exposure ........................................................................................................... 30
2.2.3 Motion sickness ................................................................................................... 31
2.2.4 Summary ............................................................................................................. 31
2.3 METHODS FOR UNDERSTANDING THE EFFECTS FROM VIBRATION EXPOSURE ...... 32
VI
2.3.1 Constant measurement methods ........................................................................ 34
2.3.1.1 Constant method ............................................................................................. 34
2.3.1.2 Method of limits ............................................................................................... 34
2.3.1.3 Method of adjustment ...................................................................................... 34
2.3.1.4 Adaptive psychological method....................................................................... 35
2.3.2 Subjective scaling methods ................................................................................. 35
2.3.2.1 Magnitude estimation method ......................................................................... 35
2.3.2.2 Difference threshold ........................................................................................ 37
2.3.2.3 Paired comparison method ............................................................................. 37
2.3.2.4 Category judgment method ............................................................................. 37
2.3.2.5 Borg scale ....................................................................................................... 38
2.3.2.6 Continuous judgment method ......................................................................... 38
2.3.3 Discussion of the methods .................................................................................. 39
2.3.4 Summary ............................................................................................................. 41
2.4 METHODS FOR EVALUATING THE EFFECTS OF VIBRATION EXPOSURE .................. 41
2.4.1 ISO 2631-1 and BS 6841 methods ..................................................................... 42
2.4.1.1 Root mean square (r.m.s.) .............................................................................. 42
2.4.1.2 Root mean quad (r.m.q.) ................................................................................. 43
2.4.1.3 Vibration dose value (VDV) ............................................................................. 43
2.4.1.4 SEAT value ..................................................................................................... 43
2.4.2 Other methods..................................................................................................... 44
2.4.2.1 Vibration greatness ......................................................................................... 44
2.4.2.2 Jerk .................................................................................................................. 44
2.4.2.3 Spinal response method of ISO 2631-5 for evaluating effects on multiple
shocks……………………………………………………………………………………………45
2.4.2.4 VDI 2057 ......................................................................................................... 45
2.4.2.5 Averaged Absorbed Power (AAP)................................................................... 46
2.4.3 Discussion of the methods .................................................................................. 46
2.4.4 Summary ............................................................................................................. 47
2.5 MEASURING VIBRATION ................................................................................... 47
2.5.1 Requirements ...................................................................................................... 48
2.5.2 Commercial equipment ....................................................................................... 49
2.5.3 Measurement procedure ..................................................................................... 50
2.5.4 Variation relating to the measurement results .................................................... 51
2.5.5 Summary ............................................................................................................. 52
2.6 STANDARDS AND GUIDELINES .......................................................................... 52
2.6.1 ISO 2631-1 (1997)............................................................................................... 53
2.6.1.1 Background ..................................................................................................... 53
2.6.1.2 Content of the standard ................................................................................... 54
VII
2.6.1.3 Interpretation of ISO 2631-1............................................................................ 59
2.6.1.4 Important changes to earlier versions ............................................................. 63
2.6.2 BS 6841 (1987) ................................................................................................... 63
2.6.2.1 History ............................................................................................................. 63
2.6.2.2 Content of the standard ................................................................................... 63
2.6.2.3 Interpretation of BS 6841 ................................................................................ 65
2.6.3 Comparison of ISO 2631-1 and BS 6841 standards ........................................... 67
2.6.4 Additional standards ............................................................................................ 69
2.6.4.1 ISO 10326 ....................................................................................................... 69
2.6.4.2 ISO 8041 ......................................................................................................... 69
2.6.5 European legislation ............................................................................................ 69
2.6.6 Guidelines for assessing discomfort from vibration ............................................ 70
2.6.7 Summary ............................................................................................................. 70
2.7 CONCLUSIONS ............................................................................................... 71
3 CHAPTER 3 – PLANNING AND SELECTING GENERAL METHODS FOR

VALIDATING AND IMPROVING THE ISO 2631-1 METHOD .......................... 72
3.1 INTRODUCTION ............................................................................................... 72

3.2 RESEARCH PLAN AND GOALS OF THE THESIS .................................................... 72
3.2.1 Goals for developing and validating equipment .................................................. 73
3.2.2 Goals of the 12-axis field measurements ............................................................ 74
3.2.3 Goals of the trials ................................................................................................ 74
3.2.3.1 Laboratory trial ................................................................................................ 74
3.2.3.2 Field trial .......................................................................................................... 74
3.3 SELECTING GENERAL METHODS FOR CONDUCTING THE RESEARCH .................... 75
3.3.1 Calculating vibration magnitudes ........................................................................ 75
3.3.2 Acquiring and calculating subjective judgments ................................................. 75
3.3.3 Statistical analysis ............................................................................................... 76
3.4 CONCLUSIONS ............................................................................................... 76
4 CHAPTER 4 - DEVELOPING AND VALIDATING EQUIPMENT FOR

MEASURING 12-AXIS VIBRATION IN A FIELD AND LABORATORY .......... 77
4.1 INTRODUCTION ............................................................................................... 77

4.2 GOALS OF THE DEVELOPMENT OF EQUIPMENT .................................................. 77
4.3 DEVELOPING AND VALIDATING MEASUREMENT EQUIPMENT ................................ 78
4.3.1 Components and methods for acquiring and detecting translational and rotational
axes………………………………………………………………………………………………… 78
4.3.1.1 Acceleration MEMS sensors ........................................................................... 78
VIII
4.3.1.2 Data acquisition ............................................................................................... 79
4.3.1.3 Accuracy of an acceleration signal.................................................................. 79
4.3.1.4 Method for calculating rotational accelerations ............................................... 80
4.3.2 Prototype I: Experiment for validating sensor configuration for measuring 12-axis
vibration……………………………………………………………………………………………. 81
4.3.2.1 Purpose ........................................................................................................... 81
4.3.2.2 Technical details.............................................................................................. 81
4.3.2.3 Validation of the accuracy of the equipment ................................................... 82
4.3.2.4 Summary ......................................................................................................... 85
4.3.3 Prototype II: Experiment for validating equipment for field measurements ........ 86
4.3.3.1 Purpose ........................................................................................................... 86
4.3.3.4 Summary ......................................................................................................... 89
4.3.4 Commercial version............................................................................................. 89
4.3.4.1 Purpose ........................................................................................................... 89
4.3.4.3 Improvements for the second prototype ......................................................... 91
4.3.4.5 Summary ......................................................................................................... 92
4.4 DEVELOPING AND VALIDATING ANALYSIS SOFTWARE ......................................... 92
4.4.1 Converting to SI-units .......................................................................................... 93
4.4.2 Frequency weighting ........................................................................................... 94
4.4.3 Root mean square values and crest factors ....................................................... 97
4.4.4 Point and overall vibration total values................................................................ 98
4.4.5 Calculating rotational axes .................................................................................. 98
4.4.6 Software development ........................................................................................ 98
4.4.7 Summary ............................................................................................................. 98
4.5 CONCLUSIONS ............................................................................................... 99
5 CHAPTER 5 - FIELD MEASUREMENTS: CONTRIBUTION OF TWELVE

AXES ACCORDING TO THE ISO 2631-1 METHOD ..................................... 100
5.1 INTRODUCTION ............................................................................................. 100

5.2 GOALS OF THE FIELD MEASUREMENTS ........................................................... 102
5.3 METHODS .................................................................................................... 103
5.3.1 Results from previous publications ................................................................... 103
5.3.2 Measurement procedure ................................................................................... 103
5.3.3 Signal processing .............................................................................................. 105
5.3.4 Vibration magnitudes ........................................................................................ 105
IX
5.3.5 Relative magnitude and contribution of the values ........................................... 106
5.3.6 Dominant axis at different frequency range ...................................................... 107
5.3.7 Effects of frequency weighting and multiplying factors to the relative contribution
of axes and locations ........................................................................................................ 107
5.3.8 Effect of the 1.4 multiplying factors for seat horizontal axes to compensate
backrest axes .................................................................................................................... 107
5.3.9 Calculating compensating factors for overall vibration total values using a
different number of axes ................................................................................................... 108
5.4 RESULTS ..................................................................................................... 108
5.4.1 Results from the previous publications ............................................................. 108
5.4.2 Comparison of vibration magnitudes on different machines ............................. 108
5.4.3 Applicability of the r.m.s. method ...................................................................... 111
5.4.4 Relative magnitude and contribution of point vibration total values .................. 112
5.4.5 Relative magnitude and contribution of individual axes .................................... 115
5.4.6 Effect of frequency weighting to the contribution of axes.................................. 119
5.4.7 Effect of multiplying factors to the contribution of axes ..................................... 122
5.4.8 The dominant axis at different 1/3 octave frequencies ..................................... 125
5.4.9 Compensating factors for point vibration total values ....................................... 127
5.5 DISCUSSION ................................................................................................. 129
5.5.1 Relative contribution of locations and axes ....................................................... 130
5.5.2 Applicability of the r.m.s. method ...................................................................... 130
5.5.3 Effects of frequency weighting .......................................................................... 131
5.5.4 Effect of multiplying factors ............................................................................... 132
5.5.5 Need for measuring additional axes.................................................................. 133
5.5.6 Approximating point vibration total values using multiplying factors ................. 133
5.5.7 Differences and similarities of new and previous measurements ..................... 134
5.5.8 Interpretation of the standard method for field measurements ......................... 134
5.5.9 Scope and limitations of the results .................................................................. 135
5.5.10 Suggestion for further work ............................................................................... 135
5.6 CONCLUSIONS ............................................................................................. 135
6 CHAPTER 6 - LABORATORY TRIAL PART I: DESCRIPTION AND

VALIDATION OF TRIAL PROCEDURE, METHODS AND RESULTS .......... 137
6.1 INTRODUCTION ............................................................................................. 137

6.2 GOALS OF THE TRIAL .................................................................................... 138
6.3 METHODS .................................................................................................... 139
6.3.1 Test environment............................................................................................... 139
6.3.1.1 Multi-axis shaker ........................................................................................... 139
6.3.1.2 Sensor setup and data acquisition ................................................................ 140
X
6.3.1.3 Stimuli ............................................................................................................ 141
6.3.2 Judgment method.............................................................................................. 146
6.3.3 Subjects ............................................................................................................. 146
6.3.4 Study procedure ................................................................................................ 147
6.3.6 Effect of sequence type and order to judgments .............................................. 148
6.3.7 Effect of subject characteristics to judgments ................................................... 149
6.4 RESULTS ..................................................................................................... 149
6.4.2 Judgment style and averaging period ............................................................... 152
6.4.3 Judgment values ............................................................................................... 154
6.4.4 Effect of sequence type and order to judgments .............................................. 155
6.4.5 Effect of subject characteristics to judgments ................................................... 156
6.5 DISCUSSION ................................................................................................. 156
6.5.1 Vibration stimuli ................................................................................................. 156
6.5.2 Judgment style and values ................................................................................ 157
6.5.3 Effects of procedure and method for results ..................................................... 157
6.5.4 Scope of the results .......................................................................................... 157
6.6 CONCLUSIONS ............................................................................................. 159
7 CHAPTER 7 - LABORATORY TRIAL PART II: VALIDATING THE ISO

2631-1 METHOD ............................................................................................ 160
7.1 INTRODUCTION ............................................................................................. 160

7.2 GOALS OF THE TRIAL PART II ......................................................................... 162
7.3 METHODS .................................................................................................... 163
7.3.2 Selecting combinations of axes for calculating overall vibration total values ... 163
7.3.3 Judgment method.............................................................................................. 165
7.3.4 Correlation analyses ......................................................................................... 165
7.4 RESULTS ..................................................................................................... 165
7.4.1 Correlation between vibration magnitudes and judgments ............................... 165
7.4.2 Effect of weighting and averaging methods ...................................................... 169
7.5 DISCUSSION ................................................................................................. 170
7.5.1 Correlation between vibration and discomfort................................................... 170
7.5.2 Effect of frequency weighting ............................................................................ 170
7.5.3 Effect of multiplying factors ............................................................................... 171
7.5.4 Effect of emphasising shocks............................................................................ 174
7.5.5 Optimal usage of the standard method ............................................................. 174
7.5.6 Comparison to previous studies ........................................................................ 175
XI
7.5.7 Scope and limitations of the results .................................................................. 175
7.5.8 Further work ...................................................................................................... 176
7.6 CONCLUSIONS ............................................................................................. 176
8 CHAPTER 8 - LABORATORY TRIAL PART III: IMPROVING THE ISO

2631-1 METHOD ............................................................................................ 178
8.1 INTRODUCTION ............................................................................................. 178

8.2 GOALS OF THE TRIAL PART III ........................................................................ 179
8.3 METHODS .................................................................................................... 179
8.3.1 Finding optimal combination of multiplying factors ........................................... 179
8.3.1.1 Brute force ..................................................................................................... 179
8.3.1.2 Multiple linear regression model ................................................................... 181
8.3.2 Visual evaluation of results ............................................................................... 182
8.3.3 Validation of results ........................................................................................... 182
8.4 RESULTS ..................................................................................................... 182
8.4.1 Best combination of multiplying factors ............................................................. 182
8.4.1.1 Brute force ..................................................................................................... 182
8.4.1.2 Linear regression models: selecting independent and dependent variables 184
8.4.1.3 Regression model 1 ...................................................................................... 186
8.4.1.4 Regression model 2 ...................................................................................... 187
8.4.2 Visual evaluation of clustering of multiplying factors......................................... 188
8.4.3 Improvement in correlation for new multiplying factors ..................................... 191
8.4.4 Validating results ............................................................................................... 192
8.4.4.1 Correlation for two random groups................................................................ 192
8.4.4.2 Multiplying factors from other studies............................................................ 193
8.4.4.3 Using new factors for data from other studies .............................................. 194
8.5 DISCUSSION ................................................................................................. 196
8.5.1 Improvement in correlation using the new multiplying factors .......................... 197
8.5.2 Confidence in the new multiplying factors ......................................................... 197
8.5.3 Comparison of results from different methods .................................................. 198
8.5.4 Applicability of the standard method ................................................................. 199
8.5.5 Scope and limitations of the new model ........................................................... 199
8.6 CONCLUSIONS ............................................................................................. 200
9 CHAPTER 9 - FIELD TRIAL: VALIDATING THE ISO 2631-1 METHOD

AND RESULTS FROM THE LABORATORY TRIAL ..................................... 201
9.1 INTRODUCTION ............................................................................................. 201

9.2 GOALS OF THE FIELD TRIAL ........................................................................... 202
XII
9.3 METHODS .................................................................................................... 203
9.3.1 Measurements................................................................................................... 203
9.3.1.1 Field environment .......................................................................................... 203
9.3.1.2 Test subjects ................................................................................................. 205
9.3.1.3 Measurement setup ...................................................................................... 206
9.3.1.4 Equipment ..................................................................................................... 208
9.3.1.5 Trial procedure .............................................................................................. 209
9.3.2 Analyses ............................................................................................................ 210
9.3.2.1 Vibration magnitudes and frequency characteristics of the legs .................. 210
9.3.2.2 Effect of additional axes to overall vibration total value ................................ 211
9.3.2.3 Analysis of judgment style ............................................................................. 211
9.3.2.4 Selecting best averaging window size and delay.......................................... 211
9.3.2.5 Normalising the judgments ............................................................................ 214
9.3.2.6 Correlation between vibration magnitudes and discomfort ........................... 214
9.3.2.7 Optimising new multiplying factors ................................................................ 215
9.3.2.8 Clustering of multiplying factors .................................................................... 215
9.4 RESULTS ..................................................................................................... 215
9.4.1 Vibration characteristics of the legs .................................................................. 216
9.4.1.1 Vibration magnitudes .................................................................................... 216
9.4.1.2 Frequency characteristics ............................................................................. 217
9.4.1.3 Effect of seat rotational and floor translational axes to overall vibration total
value…………………………………………………………………………………………… 218
9.4.1.4 Correlation between the seat and backrest translational axes ..................... 219
9.4.2 Characteristics of judgment response ............................................................... 220
9.4.2.1 Delay between vibration and judgment ......................................................... 220
9.4.2.2 Analysis of judgment style ............................................................................. 220
9.4.3 Best setup and data .......................................................................................... 224
9.4.3.1 Best window and overlap size and delay for vibration and judgment data ... 224
9.4.4 Valid subjects and legs for correlation analysis ................................................ 229
9.4.5 Normalising judgment data for comparison of subjects and groups ................. 229
9.4.5.1 Each subject .................................................................................................. 229
9.4.5.2 Each group .................................................................................................... 230
9.4.6 Correlation between vibration magnitudes and discomfort ............................... 231
9.4.6.1 Best standard scenario ................................................................................. 231
9.4.6.2 Optimised multiplying factors from the laboratory trial .................................. 232
9.4.6.3 New multiplying factors ................................................................................. 232
9.4.6.4 Clustering of the factors ................................................................................ 234
9.5 DISCUSSION ................................................................................................. 235
9.5.1 Vibration magnitudes and frequency characteristics of the legs ....................... 235
XIII
9.5.2 The effect of the additional axes ....................................................................... 236
9.5.3 Analysing different judgment styles and averaging the response ..................... 236
9.5.4 Correlation between vibration magnitudes and discomfort ............................... 237
9.5.5 Can discomfort be predicted practically using the standard or an optimised
method…………………………………………………………………………………………….238
9.5.6 Recommended methods to evaluate discomfort from whole-body vibration in
field……………………………………………………………………………………………….. 239
9.5.6.1 What worked ................................................................................................. 239
9.5.6.2 What did not work.......................................................................................... 240
9.5.7 Limitations of the results and model.................................................................. 240
9.5.8 Further work ...................................................................................................... 241
9.6 CONCLUSIONS ............................................................................................. 242
10 CHAPTER 10: GENERAL DISCUSSION ............................................. 244
10.1 INTRODUCTION ............................................................................................. 244

10.2 BENEFITS OF UNDERSTANDING DISCOMFORT FROM VIBRATION EXPOSURE ....... 244
10.3 PROBLEMS RELATING TO THE CURRENT ISO 2631-1 STANDARD ...................... 245
10.4 PREVIOUS RESEARCH AND RESULTS .............................................................. 246
10.5 MEASURING VIBRATION AND SUBJECTIVE DISCOMFORT ................................... 247
10.5.1 Measuring 12-axis vibration .............................................................................. 248
10.5.2 Acquiring judgment data ................................................................................... 248
10.6 RELATIVE CONTRIBUTION OF THE TWELVE AXES IN THE FIELD .......................... 249
10.7 USING THE STANDARD METHOD TO PREDICT DISCOMFORT .............................. 249
10.7.1 laboratory trial.................................................................................................... 250
10.7.2 field trial ............................................................................................................. 250
10.8 IMPROVING THE STANDARD METHOD .............................................................. 251
10.8.1 Role of the axes ................................................................................................ 251
10.8.2 Role of the frequency weighting curves ............................................................ 252
10.8.3 Role of the averaging method ........................................................................... 252
10.8.4 Role of the multiplying factors ........................................................................... 253
10.9 IMPLICATIONS OF THE FINDINGS FOR IMPROVING THE STANDARD METHOD ........ 253
10.10 AN IMPROVED METHOD FOR PREDICTING DISCOMFORT FROM WHOLE-BODY
VIBRATION .............................................................................................................. 255
10.11 CONTRIBUTION OF THE THESIS WORK ......................................................... 256
11 CHAPTER 11: LIMITATIONS AND FUTURE WORK .......................... 258
11.1 LIMITATIONS ................................................................................................. 258

11.1.1 Limitations of the standard method ................................................................... 258
XIV
11.1.2 Limitations of the theory behind the standard and applying the standard in
practice…………………………………………………………………………………………… 259
11.1.3 Limitations of the results ................................................................................... 260
11.1.3.1 Laboratory experiments for validating the equipment ............................... 260
11.1.3.2 Field measurements .................................................................................. 260
11.1.3.3 The laboratory trial .................................................................................... 260
11.1.3.4 The field trial .............................................................................................. 261
11.2 FURTHER WORK ........................................................................................... 262
12 CHAPTER 12 – CONCLUSIONS ......................................................... 264
12.1 CONCLUSIONS ............................................................................................. 264
13 REFERENCES...................................................................................... 269
XV
List of Tables
TABLE 1. PUBLICATIONS BASED ON THE RESULTS OF THE THESIS....................................... 8
TABLE 2. NUMBER OF PUBLICATIONS FOUND FROM DATABASE REFERRING TO WHOLE-BODY
VIBRATION AND DISCOMFORT. ENGINEERINGVILLAGE2 DATABASE INCLUDES
COMPENDEX, INSPEC, NTIS, GEOBASE, REFEREX AND OTHER ENGINEERING
DATABASES. INFORMAWORLD IS SITE HOSTING JOURNALS AND EBOOKS PUBLISHED BY
TAYLOR & FRANCIS, ROUTLEDGE, PSYCHOLOGY PRESS AND INFORMA HEALTHCARE.
SCOPUS IS THE LARGEST ABSTRACT AND CITATION DATABASE OF RESEARCH
LITERATURE AND QUALITY WEB SOURCES. SPRINGERLINK IS AN INTEGRATED FULL-TEXT
DATABASE FOR JOURNALS AND BOOKS PUBLISHED BY SPRINGER. SEARCH WAS MADE
USING KEYWORDS, ABSTRACTS AND TITLES. ........................................................... 14
TABLE 3. SUMMARY OF FIELD STUDIES REGARDING SUGGESTED METHODS FOR

MANIPULATING VIBRATION MEASUREMENTS TO PREDICT DISCOMFORT (INCLUDING
STUDIES BEFORE AND AFTER THE STANDARD). ........................................................ 26
TABLE 4. PUBLICATION DATES AND CODE NAMES OF THE MAIN VIBRATION STANDARDS. ..... 53
TABLE 5. FREQUENCY WEIGHTINGS AND MULTIPLYING FACTORS DEFINED IN ISO 2631-1 FOR
ANALYSING 12-AXIS MEASUREMENTS FOR A SEAT PERSON FOR EVALUATING
DISCOMFORT. ....................................................................................................... 55
TABLE 6. APPROXIMATE INDICATIONS OF LIKELY REACTIONS TO VARIOUS MAGNITUDES

OF OVERALL VIBRATION TOTAL VALUES IN PUBLIC TRANSPORT (ISO 2631-1 1997).
........................................................................................................................... 58
TABLE 7. DIFFERENT APPROACHES USED IN THE THESIS TO CALCULATE VIBRATION
MAGNITUDES FOR EACH AXIS. FREQUENCY WEIGHTING AND MULTIPLYING FACTORS ARE
APPLIED BASED ON THE ISO 2631-1 STANDARD. AVERAGING METHODS ARE APPLIED
BASED ON THE ISO 2631-1 AND BS 6841 STANDARDS. POINT AND OVERALL VIBRATION
TOTAL VALUES ARE CALCULATED FOR ALL SETUPS. ................................................. 75
TABLE 8. TECHNICAL SPECIFICATIONS OF THE MEMS SENSOR SELECTED FOR THE 12-AXIS
MEASUREMENT SYSTEM. ....................................................................................... 79
TABLE 9. SUMMARY OF THE TEST STIMULI CONDUCTED FOR EVALUATING THE ACCURACY OF
THE DEVELOPED SENSOR CONFIGURATION. AT LEAST TWO FREQUENCIES AND
AMPLITUDES WERE USED FOR ALL POSSIBLE COMBINATIONS OF AXES (ALL SIX AXES
SEPARATELY AND COMBINATION OF ONE TRANSLATIONAL AND ROTATIONAL AXIS). .... 83
TABLE 10. THE THEORETICAL SENSITIVITY (I.E. SMALLEST DETECTABLE CHANGE IN

VIBRATION AMPLITUDE) FOR DIFFERENT AMPLITUDE RANGES OF THE SELECTED MEMS
SENSOR. .............................................................................................................. 86
TABLE 11. THE SENSITIVITY OF PROTOTYPE II FOR ALL SENSORS AND AXES BASED ON THE
GRAVITY TEST. ...................................................................................................... 88
TABLE 12. THE SENSITIVITIES OF THE COMMERCIAL V6SP SENSOR FOR 1.5 AND 3.0 G
VERSIONS USING NATIONAL INSTRUMENTS NI-DAQ SYSTEM WITH 5 V INPUT. ........... 92
TABLE 13. W EIGHING VALUES FOR THE STANDARD AND DEVELOPED WEIGHTING CURVES
1
AND ERRORS (%) BETWEEN THEM AT EACH /3 OCTAVE FREQUENCY. ........................ 96
TABLE 14. CALCULATED R.M.S. VALUES FOR THE STANDARD AND DEVELOPED FREQUENCY
WEIGHTINGS USING GUIDED REFERENCE FREQUENCIES. .......................................... 97
TABLE 15. VALIDATING THE R.M.S. AND CSF CALCULATIONS USING A REFERENCE SIGNAL. 97
XVI
TABLE 16. REFERENCE VALUES TO VALIDATE IF PROGRAMMED ALGORITHMS TO CALCULATE
POINT AND OVERALL VIBRATION TOTAL VALUES WORK. BOTH VALUES ARE CALCULATED
USING A VECTOR SUM PRINCIPLE (EQUATION 13). ................................................... 98
TABLE 17. THE MEASURED MACHINES AND INFORMATION FROM THE NEW FIELD
MEASUREMENTS CONDUCTED IN THIS THESIS. EACH MACHINE AND WORK PHASE WAS
MEASURED IN AN AUTHENTIC ENVIRONMENT, WHERE THE OPERATOR WORKED
NORMALLY. ......................................................................................................... 105
TABLE 18. THE FREQUENCY WEIGHTED POINT VIBRATION TOTAL VALUES (M/S2) FOR SEAT
TRANSLATIONAL AXES WITH 1.0 MULTIPLYING FACTORS (AS1.0), BACKREST
TRANSLATIONAL AXES (AB), FLOOR TRANSLATIONAL AXES (AF) AND SEAT ROTATIONAL
AXES (AR) WITH RESPECTIVE MULTIPLYING FACTORS. THE FREQUENCY WEIGHTED
2
OVERALL VIBRATION TOTAL VALUES (M/S ) FOR SEAT TRANSLATIONAL AXES WITH 1.4
MULTIPLYING FACTORS (AS1.4), SEAT AND BACKREST TRANSLATIONAL AXES (ASB), SEAT,
BACKREST AND FLOOR TRANSLATIONAL AXES (ASBF) AND ALL AXES (AV) WITH
RESPECTIVE MULTIPLYING FACTORS FOR ALL MEASUREMENTS. RELATIVE DIFFERENCES
(%) ARE COMPARED TO THE OVERALL VIBRATION TOTAL VALUE USING ALL AXES (AV).
......................................................................................................................... 110
TABLE 19. THE FREQUENCY WEIGHTED R.M.S. VALUES (M/S2 OR RAD/S2) WITHOUT
MULTIPLYING FACTORS FOR SEAT TRANSLATIONAL AXES (XS, YS AND ZS), SEAT
ROTATIONAL AXES (ROLL, PITCH AND YAW ), BACKREST TRANSLATIONAL AXES (XB, YB
AND ZB) AND FLOOR TRANSLATIONAL AXES (XF, YF AND ZF) FOR ALL MEASUREMENTS.
......................................................................................................................... 111
TABLE 20. THE CREST FACTOR VALUES FOR FREQUENCY WEIGHTED SEAT TRANSLATIONAL
AXES (XS, YS AND ZS), SEAT ROTATIONAL AXES (ROLL, PITCH AND YAW ), BACKREST
TRANSLATIONAL AXES (XB, YB AND ZB) AND FLOOR TRANSLATIONAL AXES (XF, YF AND
ZF). VALUES EXCEEDING THE THRESHOLD LIMIT 9 OF ISO 2631-1 ARE HIGHLIGHTED.
......................................................................................................................... 112
TABLE 21. THE RELATIVE MAGNITUDES (%) OF POINT VIBRATION TOTAL VALUES FOR SEAT
TRANSLATIONAL AXES (AS1.0), BACKREST TRANSLATIONAL AXES (AB), FLOOR
TRANSLATIONAL AXES (AF) AND SEAT ROTATIONAL AXES (AR) WITH RESPECTIVE
MULTIPLYING FACTORS. THE LOCATION WHICH WAS THE HIGHEST IS TREATED AS THE
MOST IMPORTANT POINT VIBRATION TOTAL VALUE (I.E. 100 %). THE REST OF THE AXES
ARE COMPARED TO THE MOST IMPORTANT POINT VIBRATION TOTAL VALUE (EMBOLDEN).
UNDERLINED VALUES ARE BELOW 25 % OF THE HIGHEST VALUE, THUS COULD BE
NEGLECTED BASED ON THE STANDARD GUIDANCE. ................................................ 113
TABLE 22. THE RELATIVE CONTRIBUTION (%) OF FREQUENCY WEIGHTED POINT VIBRATION
2
TOTAL VALUES (M/S ) FOR SEAT TRANSLATIONAL AXES (AS1.0), BACKREST
TRANSLATIONAL AXES (AB), FLOOR TRANSLATIONAL AXES (AF) AND SEAT ROTATIONAL
AXES (AR) WITH RESPECTIVE MULTIPLYING FACTORS TO THE OVERALL VIBRATION TOTAL
VALUE OF ALL AXES. ............................................................................................ 114
TABLE 23. THE RELATIVE MAGNITUDES (%) OF THE FREQUENCY WEIGHTED SEAT
TRANSLATIONAL AXES (XS, YS AND ZS), SEAT ROTATIONAL AXES (ROLL, PITCH AND
YAW ), BACKREST TRANSLATIONAL AXES (XB, YB AND ZB) AND FLOOR TRANSLATIONAL
AXES (XF, YF AND ZF) WITH MULTIPLYING FACTORS. THE AXIS WHICH WAS THE
HIGHEST IS TREATED AS THE MOST IMPORTANT AXIS (I.E. 100 %). THE REST OF THE
AXES ARE COMPARED TO THE MOST IMPORTANT AXIS (EMBOLDEN). ........................ 116
TABLE 24. THE RELATIVE CONTRIBUTION (%) OF FREQUENCY WEIGHTED SEAT

XVII
AXES (XF, YF AND ZF) TO THE OVERALL VIBRATION TOTAL VALUE WITH RESPECTIVE
MULTIPLYING FACTORS. ....................................................................................... 118
TABLE 25. THE RELATIVE DIFFERENCE RATIOS (%) OF R.M.S. VALUES WITH AND WITHOUT
FREQUENCY WEIGHTING FOR SEAT TRANSLATIONAL AXES (XS, YS AND ZS), SEAT
ROTATIONAL AXES (ROLL, PITCH AND YAW ), BACKREST TRANSLATIONAL AXES (XB, YB
AND ZB) AND FLOOR TRANSLATIONAL AXES (XF, YF AND ZF), AND FOR OVERALL
VIBRATION TOTAL VALUES FOR SEAT TRANSLATIONAL AXES WITH 1.0 (AS1.0), SEAT AND
BACKREST AXES (ASB), ALL TRANSLATIONAL AXES (AFSB), AND TO ALL TWELVE AXES (AV)
THE PERCENTAGE IS RATIO OF THE WEIGHTED VALUE TO THE UNWEIGHTED VALUE. 120
TABLE 26. OVERALL VIBRATION TOTAL VALUES OF FREQUENCY WEIGHTED R.M.S. VALUES
(M/S2) WITH AND WITHOUT MULTIPLYING FACTORS, AND RELATIVE DIFFERENCES OF
VALUES WITH FACTORS TO WITHOUT FACTORS. ..................................................... 123
TABLE 27. COMPARISON OF OVERALL VIBRATION TOTAL VALUES (M/S2) OF SEAT

TRANSLATIONAL AXES WITH 1.4 MULTIPLYING FACTORS (AS1.4) AND WITH SEAT AND
BACKREST TRANSLATIONAL AXES (ASB). AND CALCULATION OF FACTOR WHICH WOULD
EQUAL AS AND ASB. DIFFERENCE (%) IS AS1.4 TO ASB. ............................................... 124
TABLE 28. COMPENSATING FACTORS FOR DIFFERENT COMBINATIONS OF POINT VIBRATION

TOTAL VALUES TO COMPENSATE THE EFFECTS OF OTHER POINT VIBRATION TOTAL
VALUES. THE FACTOR IS USED TO MULTIPLY THE CALCULATED OVERALL VIBRATION
TOTAL VALUE. ..................................................................................................... 129
TABLE 29. MAIN STIMULI USED IN THE TRIAL TO REPRESENT VIBRATION FROM THE FIELD.
EACH STIMULUS WAS BASED ON A SELECTED DATA SAMPLE OF THE FIELD
MEASUREMENTS IN CHAPTER 5. EACH MAIN STIMULUS LASTED 15 SECONDS. ......... 141
TABLE 30. STIMULI AND CONTROL FACTORS FOR THE TEST BENCH FOR ALL STIMULI
VARIATIONS. ....................................................................................................... 145
TABLE 31. INFORMATION ON THE TEST SUBJECTS USED IN THE TRIAL. THE STIMULI WERE
RANDOMISED AND THREE SEQUENCES (A, B AND C) WERE CREATED. THE ORDER OF
SEQUENCE FOR EACH SUBJECT WAS CHANGED. .................................................... 147
TABLE 32. THE FREQUENCY WEIGHTED R.M.S. VALUES (M/S2) WITH RESPECTIVE
MULTIPLYING FACTORS OF EACH STIMULUS FOR EACH MEASURED AXIS AND OVERALL
VIBRATION TOTAL VALUE (OVTV) USING SEAT TRANSLATIONAL AND ROTATIONAL AXES
AND BACKREST TRANSLATIONAL AXES. ................................................................. 150
TABLE 33. AVERAGED JUDGMENT VALUES FOR EACH SEQUENCE AND TEST SUBJECT, AND
THE ORDER WHICH THE SEQUENCES WERE RUN FOR EACH SUBJECT. ..................... 154
TABLE 34. MANN-W HITNEY U TEST FOR DIFFERENCES IN JUDGMENT BETWEEN MEN AND
WOMEN. ............................................................................................................. 156
TABLE 35. PRACTICALLY REALISABLE SCENARIOS FOR CALCULATING AN OVERALL VIBRATION

TOTAL VALUE BASED ON THE ISO 2631-1 STANDARD WHEN USING THE SEAT,
BACKREST AND FLOOR TRANSLATIONAL AXES. ...................................................... 164
TABLE 36. THE CORRELATION (R2) BETWEEN DISCOMFORT JUDGMENTS OF INDIVIDUAL

SUBJECTS AND FREQUENCY WEIGHTED VIBRATION MAGNITUDES USING THE R.M.S.
METHOD WITH RESPECTIVE MULTIPLYING FACTORS FOR EACH SCENARIO (THE BEST
SCENARIO FOR EACH SUBJECT IS EMBOLDEN AND UNDERLINED). ............................ 167
TABLE 37. THE CORRELATION (R2) BETWEEN DISCOMFORT JUDGMENTS AND FREQUENCY
WEIGHTED POINT VIBRATION TOTAL VALUES WITH RESPECTIVE MULTIPLYING FACTORS
AND THE DOMINANT (I.E. LARGEST) POINT VIBRATION TOTAL VALUE FOR EACH TEST
SUBJECT (BEST CORRELATIONS ARE EMBOLDEN AND UNDERLINED). ....................... 168
XVIII
TABLE 38. COMPARISON OF CORRELATIONS (R2) FOR USING THE R.M.S. AND R.M.Q.
AVERAGING METHODS FOR THE FREQUENCY WEIGHTING WITH THE MULTIPLYING
FACTORS, THE FREQUENCY WEIGHTING WITHOUT THE MULTIPLYING FACTORS AND
WITHOUT THE FREQUENCY WEIGHTING AND THE MULTIPLYING FACTORS. ................ 169
TABLE 39. THE RANGE AND STEPS OF MULTIPLYING FACTORS FOR BRUTE FORCE SEARCH.
......................................................................................................................... 181
TABLE 40. THE TOP TEN COMBINATIONS OF MULTIPLYING FACTORS FOR ALL MEASURED NINE
AXES. ................................................................................................................. 183
TABLE 41. THE CORRELATIONS (PEARSON R) AND P-VALUES FOR ALL INDEPENDENT
VARIABLES. ........................................................................................................ 184
TABLE 42. THE SELECTED DEPENDENT AND INDEPENDENT VARIABLES USED FOR
CALCULATING THE REGRESSION MODEL. THE FREQUENCY WEIGHTED R.M.S. VALUES
(WITHOUT MULTIPLYING FACTORS) FOR SEAT TRANSLATIONAL AXES ARE USED AS
INDEPENDENT VARIABLES AND AVERAGED JUDGMENTS OF SUBJECTS FOR EACH STIMULI
ARE USED AS DEPENDENT VALUES. ...................................................................... 185
TABLE 43. MULTIPLE LINEAR REGRESSION MODEL PARAMETERS FOR MODEL 1. .............. 186
TABLE 44. SUMMARY FOR REGRESSION MODEL 1 (PEARSON)........................................ 186
TABLE 45. PARAMETER TRANSFORMATION TO COMPLY WITH THE STANDARD MULTIPLYING
FACTORS, WHERE VERTICAL AXIS HAS 1.0 FACTOR. ............................................... 186
TABLE 46. MULTIPLE LINEAR REGRESSION MODEL PARAMETERS FOR MODEL 2. .............. 187
TABLE 47. MODEL SUMMARY. ...................................................................................... 187
TABLE 48. PARAMETER TRANSFORMATION TO COMPLY WITH THE STANDARD MULTIPLYING
FACTORS, WHERE VERTICAL AXIS HAS 1.0 FACTOR. ............................................... 188
TABLE 49. COMPARISON BETWEEN CORRELATIONS OF THE BEST STANDARD SCENARIO AND
SCENARIOS WITH OPTIMISED MULTIPLYING FACTORS FOR THE SEAT TRANSLATIONAL
AXES (SCENARIO 8: SEAT FORE-AND-AFT 2.7, LATERAL 1.8 AND VERTICAL 1.0,
SCENARIO 9: SEAT FORE-AND-AFT 4.1, LATERAL 2.3 AND VERTICAL 1.0, SCENARIO 10:
SEAT FORE-AND-AFT 2.78, LATERAL 1.68 AND VERTICAL 1.0). SCENARIOS 8 AND 10
USE ROOT-SUM-OF-SQUARES METHOD FOR COMBINING THE AXES, WHILE SCENARIO 9
USES DIRECT SUMMING. ...................................................................................... 192
TABLE 50. PARTICIPANTS DIVIDED INTO TWO GROUPS. .................................................. 193

TABLE 51. COMPARISON BETWEEN THE BEST STANDARD SCENARIOS (6 AND 7) AND BEST
NEW SCENARIO (8) FOR GROUPS 1, 2 AND ALL. SCENARIO 6 INCLUDES SEAT
TRANSLATIONAL (WITH 1.4 MULTIPLYING FACTORS FOR FORE-AND-AFT AND LATERAL
AXES) AND ROTATIONAL AXES (WITH RESPECTIVE MULTIPLYING FACTORS), SCENARIO 7
INCLUDES SEAT TRANSLATIONAL AND ROTATIONAL AXES AND BACKREST
TRANSLATIONAL AXES WITH RESPECTIVE MULTIPLYING FACTORS, AND SCENARIO 8
INCLUDES SEAT TRANSLATIONAL AXES WITH 2.7, 1.8 AND 1.0 MULTIPLYING FACTORS (X,
Y AND Z). ............................................................................................................ 193
TABLE 52. SUMMARY OF THE WEIGHTED VIBRATION MAGNITUDES USED BY MANSFIELD AND
MAEDA (2006) (ADAPTED FROM THE SOURCE). ..................................................... 195
TABLE 53. LEGS USED IN THE FIELD TRIAL. ................................................................... 203
TABLE 54. THE CHARACTERISTICS OF THE TEST SUBJECTS. GROUP 3 HAD SUBJECTS FROM
GROUPS 1 AND 2 (SAME SUBJECT ID). EACH SEAT HAD A FIXED SEAT ID NUMBER. ... 206
XIX
TABLE 55. SETUPS FOR TESTING BEST PARAMETERS FOR COMPARING CORRELATION
BETWEEN VIBRATION AND JUDGMENT VALUES. ...................................................... 213
TABLE 56. PRACTICALLY REALISABLE SCENARIOS BASED ON THE ISO 2631-1 STANDARD
WHEN USING THE SEAT AND BACKREST TRANSLATIONAL AXES. ............................... 215
TABLE 57. FREQUENCY WEIGHTED R.M.S. VALUES (M/S2) WITHOUT THE MULTIPLYING
FACTORS, AND STANDARD DEVIATIONS FOR ALL LEGS (WITHOUT MULTIPLYING
FACTORS) AVERAGED FROM SUBJECTS FROM GROUP 1. ........................................ 216
TABLE 58. SUMMARY OF ALL LEGS AND EVALUATED JUDGMENT STYLES. ........................ 224
TABLE 59. CORRELATION (SPEARMAN R2) FOR ALL SETUPS FOR EACH SUBJECT USING
STANDARD SCENARIO 2 (THE SEAT TRANSLATIONAL AXES WITH 1.4 MULTIPLYING
FACTORS FOR THE HORIZONTAL AXES). GROUP 3 HAD SUBJECTS ALSO FROM THE
PREVIOUS GROUPS 1 AND 2. ................................................................................ 228
TABLE 60. CORRELATION (SPEARMAN R2) FOR ALL SCENARIOS FOR EACH SUBJECT. GROUP
3 HAD SUBJECTS ALSO FROM PREVIOUS GROUPS. ................................................. 231
TABLE 61. CORRELATION (SPEARMAN R2) FOR ALL SCENARIOS FOR EACH GROUP .......... 232
TABLE 62. OPTIMISED MULTIPLYING FACTORS AND CORRELATION (SPEARMAN) FOR EACH
SUBJECT USING THE BRUTE FORCE METHOD. SUBJECTS 5 AND 6 FROM GROUP 1, AND
10 FROM GROUP 2 WERE DISMISSED FROM THE ANALYSES. ................................... 233
TABLE 63. OPTIMISED MULTIPLYING FACTORS AND CORRELATION (SPEARMAN) FOR EACH
GROUP USING THE BRUTE FORCE METHOD. ........................................................... 234
XX
List of Figures
FIGURE 1. ISO 2631-1 (1997) STANDARD METHOD FOR COMBINING THE FREQUENCY
WEIGHTED VIBRATION DATA FROM EACH OF THE TWELVE AXES TO A SINGLE “OVERALL
VIBRATION TOTAL VALUE”. THE R.M.Q AND VDV METHODS CAN BE USED INSTEAD OF
THE R.M.S. METHOD FOR AVERAGING THE VIBRATION DATA. ....................................... 4
FIGURE 2. THE TWELVE AXES AND LOCATIONS USED TO DETERMINE VIBRATION DISCOMFORT
FOR SEATED SUBJECTS (GRIFFIN, ET AL 1982)........................................................ 12
FIGURE 3. MEDIAN EQUIVALENT COMFORT CONTOURS OF DIFFERENT STUDIES FOR VERTICAL

AXIS AT SEAT FOR A SEATED PERSON COMPARED BETWEEN PREVIOUS STUDIES (- - -)
AND THE REFERENCE STUDY (---) (GRIFFIN, ET AL 1982). ........................................ 16
FIGURE 4. MEDIAN, 25TH AND 75TH PERCENTILE COMFORT CONTOURS FOR VERTICAL SEAT
VIBRATION (GRIFFIN, W HITHAM AND PARSONS 1982). ............................................. 16
FIGURE 5. MEDIAN, 25TH AND 75TH PERCENTILE COMFORT CONTOURS FOR FORE-AND-AFT
(LEFT) AND LATERAL (RIGHT) SEAT VIBRATION (GRIFFIN, ET AL 1982). ...................... 17
FIGURE 6. MEDIAN EQUIVALENT COMFORT CONTOURS FOR FORE-AND-AFT VIBRATION,
COMPARED BETWEEN GRIFFIN ET AL (1982) AND MIWA (- - -) (MIWA 1967)............... 17
FIGURE 7. MIWA’S STUDIES ON SEAT TRANSLATIONAL AXES USING VIBRATION GREATNESS

METHOD FOR SINUSOIDAL STIMULI (LEFT) AND RANDOM STIMULI (RIGHT) (MIWA AND
YONEKAWA 1974)................................................................................................. 18
FIGURE 8. COMFORT CONTOURS FOR THREE TRANSLATIONAL AXES OF BACK (LEFT) AND
FOOT (RIGHT) VIBRATION (PARSONS, ET AL 1982). .................................................. 19
FIGURE 9. MEDIAN, 25TH AND 75TH PERCENTILE COMFORT CONTOURS FOR ROLL (LEFT),
PITCH (MIDDLE) AND YAW (RIGHT) DIRECTIONS (PARSONS AND GRIFFIN 1982) .......... 19
FIGURE 10. COMPARISON OF BS 6841 STANDARD WEIGHTINGS TO NEWER FINDINGS.

EFFECT OF VIBRATION MAGNITUDE ON FREQUENCY WEIGHTINGS (INVERTED
EQUIVALENT COMFORT CONTOURS NORMALISED AT 2HZ FOR FORE-AND-AFT AND
LATERAL VIBRATION AND NORMALISED AT 5HZ FOR VERTICAL VIBRATION): (A) FORE-
AND-AFT, (B) LATERAL, (C) VERTICAL. A SENSATION MAGNITUDE OF 100 IS EQUIVALENT
2
TO THE DISCOMFORT PRODUCED BY 1.0 M/S RMS (FORE-AND-AFT AND LATERAL
2
VIBRATION) OR 0.5 M/S RMS (VERTICAL VIBRATION) AT 20 HZ (MORIOKA AND GRIFFIN
2006). ................................................................................................................. 23
FIGURE 11. EXAMPLES OF PSYCHOPHYSICAL METHODS FOR STUDYING COMFORT AND
PERCEPTION FROM VIBRATION EXPOSURE (ADAPTED FROM MAEDA (2005)). ............. 34
FIGURE 12. EBE’S MODEL FOR PREDICTING SEAT DISCOMFORT USING STATIC AND DYNAMIC
PROPERTIES (EBE AND GRIFFIN 2000A). ................................................................ 36
FIGURE 13. IMPROVED MODEL FOR PREDICTING SEAT DISCOMFORT USING STATIC, DYNAMIC
AND TEMPORAL FACTORS (MANSFIELD 2005). ........................................................ 37
FIGURE 14. THE DIFFERENCE OF SUBJECTIVE SCALES COMPARED TO ISO 2631-1 SCALE
AND USING CATEGORY JUDGMENT METHOD (KANEKO, ET AL 2005). ......................... 40
FIGURE 15. COMMERCIAL 12-AXIS MEASUREMENT EQUIPMENT FOR DISCOMFORT

EVALUATION BY IMV CO (CROPPED FROM A PRODUCT FLYER). ................................ 50
FIGURE 16. 12-AXIS MEASUREMENT SYSTEM INSTALLED ON A BUS SEAT ACCORDING TO

STANDARD GUIDANCE (YAMASHITA AND MAEDA 2003). ........................................... 51
XXI
FIGURE 17. AXES AND LOCATIONS WHERE VIBRATION SHOULD BE MEASURED FOR
ANALYSING DISCOMFORT OF A SEATED PERSON (ISO 1997). ................................... 54
FIGURE 18. PRINCIPAL WEIGHTING CURVES OF ISO 2631-1 FOR VERTICAL AND HORIZONTAL
AXES FOR EVALUATING COMFORT. W K IS USED FOR VERTICAL AXIS FROM SEAT AND ALL
FLOOR TRANSLATIONAL AXES. W D IS USED FOR SEAT AND BACKREST HORIZONTAL AXES
(ISO 2631-1 1997)............................................................................................... 55
FIGURE 19. W EIGHTING CURVES OF ISO 2631-1 FOR ROTATIONAL (W E) AND BACKREST
FORE-AND-AFT AXIS (W C) (ISO 2631-1 1997). ........................................................ 56
FIGURE 20. MEASUREMENT AND ANALYSIS PROCEDURE OF ISO 2631-1 STANDARD FOR
EVALUATING COMFORT FROM WHOLE-BODY VIBRATION. .......................................... 59
FIGURE 21. MEASUREMENT AND ANALYSIS PROCEDURE OF BS 6841 FOR EVALUATING

COMFORT OF SEATED PERSONS FROM WHOLE-BODY VIBRATION. ............................. 65
FIGURE 22. SEAT PAD SENSOR MEASURES ACCORDING TO ISO 10326 (1992). ................ 69
FIGURE 23. SENSOR LOCATIONS AND ORIENTATIONS FOR CALCULATING 6-AXIS VIBRATION
USING 6-AXIS ACCELEROMETERS AT POINTS A0, A1, A2 AND A4. ............................. 80
FIGURE 24. THE 6-AXIS SENSOR CONFIGURATION FOR MEASURING TRANSLATIONAL AND
ROTATIONAL AXES FROM SEAT (LEFT) AND THE MEASUREMENT SYSTEM TO ACQUIRE
MEASUREMENT DATA (RIGHT). ............................................................................... 82
FIGURE 25. JAPAN NIOSH 6-AXIS TEST BENCH (LEFT) AND INSTALLED SENSORS (RIGHT).
THE MEMS AND IMV SENSORS WERE INSTALLED TO THE SAME LOCATION TO COMPARE
RESULTS. ............................................................................................................. 82
FIGURE 26. RANDOM VIBRATION IN TIME DOMAIN (LEFT) AND FREQUENCY DOMAIN (RIGHT)
FROM THE TEST BENCH FOR MEMS AND IMV. ........................................................ 84
FIGURE 27. A TRANSFER FUNCTION ESTIMATE COMPARING MEMS TO IMV FOR VERTICAL
AXIS. SAME ANALYSIS WAS MADE FOR ALL AXES FOR EACH STIMULUS. ...................... 84
FIGURE 28. COMPARISON OF THE R.M.S. VALUES OF IMV AND MEMS FOR ALL ANALYSED
STIMULI. R.M.S. VALUES ARE CALCULATED FOR COMBINING THREE TRANSLATIONAL AND
THREE ROTATIONAL AXES BASED ON THE OVERALL VIBRATION TOTAL VALUE OF THE
ISO 2631-1 STANDARD. ........................................................................................ 85
FIGURE 29. THE 6-AXIS SEAT SENSOR AND DATA ACQUISITION SYSTEM, WHICH INCLUDES
ALSO MEMORY AND PROCESSOR FOR ACQUIRING AND CALCULATING SENSOR DATA. . 87
FIGURE 30. VALIDATION OF SENSITIVITY OF PROTOTYPE II BY USING GRAVITATION TEST FOR

A VERTICAL AXIS OF ONE OF THE SENSORS. ............................................................ 89
FIGURE 31. THE COMMERCIAL 6-AXIS SEAT SENSOR V6SP BOARD AND MECHANICAL
HOUSING. ............................................................................................................. 90
FIGURE 32. THE COMMERCIAL 3-AXIS BACKREST AND FLOOR SENSOR V3SP (LEFT:
ELECTRONIC BOARD, RIGHT: MECHANICAL HOUSING). .............................................. 91
FIGURE 33. THE ANALYSIS PROCEDURE OF THE DEVELOPED MEASUREMENT EQUIPMENT AND
SOFTWARE FOR MEASURING, CALCULATING AND EVALUATING DISCOMFORT FROM
WHOLE-BODY VIBRATION USING THE FULL METHOD OF ISO 2631-1. ......................... 93
FIGURE 34. CONVERTING THE RAW ACCELERATION DATA FROM A/D VALUES TO SI UNITS
(M/S2) (LEFT: BEFORE AND RIGHT: AFTER). .............................................................. 94
FIGURE 35. FREQUENCY WEIGHTING CURVES W K (LEFT) AND W D (RIGHT) FOR BOTH
DEVELOPED AND STANDARD’S FILTERS IN FREQUENCY DOMAIN. ............................... 95
XXII
FIGURE 36. FREQUENCY WEIGHTING CURVES W C (LEFT) AND W E (RIGHT) FOR BOTH
DEVELOPED AND STANDARD’S FILTERS IN FREQUENCY DOMAIN. ............................... 95
FIGURE 37. AN EXAMPLE OF THE EFFECT OF FREQUENCY WEIGHTING W D FOR THE

ACCELERATION DATA IN TIME DOMAIN (LEFT: BEFORE AND RIGHT: AFTER). ................ 97
FIGURE 38. THE SENSORS WERE INSTALLED TO THE FLOOR, SEAT SURFACE AND BACKREST
BASED ON THE GUIDANCE OF THE ISO 2631-1 STANDARD. THE EQUIPMENT USED IN
THE NEW FIELD MEASUREMENTS WERE BASED ON THE SECOND PROTOTYPE VERSION
(CHAPTER 4). ..................................................................................................... 104
FIGURE 39. RELATIVE CONTRIBUTION (%) OF FREQUENCY WEIGHTED POINT VIBRATION
TOTAL VALUES WITH RESPECTIVE MULTIPLYING FACTORS TO THE OVERALL VIBRATION
TOTAL VALUE. ..................................................................................................... 115
FIGURE 40. THE RELATIVE CONTRIBUTION (%) OF FREQUENCY WEIGHTED SEAT

AXES (XF, YF AND ZF) TO THE OVERALL VIBRATION TOTAL VALUE WITH RESPECTIVE
MULTIPLYING FACTORS. ....................................................................................... 119
FIGURE 41. FREQUENCY SPECTRA (DBM/HZ) OF TWELVE UNWEIGHTED AXES FOR A CAR
DRIVEN ON A COBBLESTONE ROAD (FREQUENCY RANGE FROM 0 TO 80 HZ). BLUE LINE
SHOWS THE MEAN SPECTRA OF TWO SEPARATE MEASUREMENTS (RED LINES). ....... 121
FIGURE 42. FREQUENCY SPECTRA (DBM/HZ) OF TWELVE WEIGHTED AXES FOR A CAR
DRIVEN ON A COBBLESTONE ROAD (FREQUENCY RANGE FROM 0 TO 80 HZ). BLUE LINE
SHOWS THE MEAN SPECTRA OF TWO SEPARATE MEASUREMENTS (RED LINES). ....... 121
FIGURE 43. DOMINANT AXES AT EACH 1/3 OCTAVE FREQUENCY FROM A CAR DRIVEN ON A
COBBLESTONE ROAD. THE TOP PLOT (RED) REPRESENTS FREQUENCY WEIGHTED
ACCELERATION DATA WITH MULTIPLYING FACTORS, MIDDLE PLOT (BLUE) REPRESENTS
WEIGHTED DATA WITHOUT MULTIPLYING FACTORS AND BOTTOM PLOT (BLUE)
UNWEIGHTED DATA WITHOUT MULTIPLYING FACTORS. ........................................... 125
FIGURE 44. DOMINANT AXES FROM BUS (UPPER LEFT), CAR (UPPER RIGHT) AND TRAIN
(BELOW ). THE TOP PLOT (RED) REPRESENTS FREQUENCY WEIGHTED ACCELERATION
DATA WITH MULTIPLYING FACTORS, MIDDLE PLOT (BLUE) REPRESENTS WEIGHTED DATA
WITHOUT MULTIPLYING FACTORS AND BOTTOM PLOT (BLUE) UNWEIGHTED DATA
WITHOUT FACTORS. ............................................................................................ 126
FIGURE 45. DOMINANT AXES FROM EXCAVATOR (UPPER LEFT), HARVESTER (UPPER RIGHT)
AND TRACTOR (BELOW ). THE TOP PLOT (RED) REPRESENTS FREQUENCY WEIGHTED
ACCELERATION DATA WITH MULTIPLYING FACTORS, MIDDLE PLOT (BLUE) REPRESENTS
WEIGHTED DATA WITHOUT MULTIPLYING FACTORS AND BOTTOM PLOT (BLUE)
UNWEIGHTED DATA WITHOUT FACTORS. ............................................................... 127
FIGURE 46. CORRELATION (SPEARMAN R2) BETWEEN FREQUENCY WEIGHTED OVERALL

VIBRATION TOTAL VALUE OF SEAT TRANSLATIONAL AXES (AS) (WITH 1.4 MULTIPLYING
FACTORS TO SEAT HORIZONTAL AXES IS SHOWN ON LEFT AND WITHOUT THE
MULTIPLYING FACTORS ON RIGHT) AND SEAT AND BACKREST TRANSLATIONAL AXES
(ASB) WITH RESPECTIVE MULTIPLYING FACTORS. THE VALUES ARE CALCULATED BASED
ON THE RESULTS FROM MAEDA (2004), GRIFFIN (1990) AND THE NEW FIELD
MEASUREMENTS. ................................................................................................ 128
FIGURE 47. CORRELATION (SPEARMAN R2) BETWEEN FREQUENCY WEIGHTED OVERALL

VIBRATION TOTAL VALUE OF SEAT TRANSLATIONAL AXES (AS) WITH 1.4 MULTIPLYING
FACTORS TO SEAT HORIZONTAL AXES AND SEAT, BACKREST AND FLOOR
TRANSLATIONAL AXES WITH RESPECTIVE MULTIPLYING FACTORS (AFSB) (LEFT), AND
XXIII
CORRELATION FOR AS AND OVERALL VIBRATION TOTAL VALUE USING ALL TWELVE AXES
WITH RESPECTIVE MULTIPLYING FACTORS (AV) (RIGHT). THE VALUES ARE CALCULATED
BASED ON THE RESULTS FROM MAEDA (2004), GRIFFIN (1990) AND THE NEW FIELD
MEASUREMENTS. ................................................................................................ 128
FIGURE 48. SIX-AXIS SHAKER AT LOUGHBOROUGH UNIVERSITY FOR CONDUCTING THE

MULTI-AXIS LABORATORY TRIAL. ........................................................................... 139
FIGURE 49. SHAKER DIMENSIONS AND SENSOR LAYOUT FOR THE LABORATORY TRIAL (SIDE
VIEW ). ................................................................................................................ 140
FIGURE 50. POWER SPECTRA FROM THE MULTI-AXIS TEST BENCH OF THE FLOOR
TRANSLATIONAL AXES FOR THE TRAIN STIMULUS (LEFT: FORE-AND-AFT, MIDDLE:
LATERAL AND RIGHT: VERTICAL AXIS).................................................................... 142
FIGURE 51. UNWEIGHTED POWER SPECTRA OF THE MAIN STIMULUS A (CAR) FOR EACH AXIS
MEASURED FROM THE SEAT SURFACE OF THE TEST BENCH. ................................... 142
FIGURE 52. UNWEIGHTED POWER SPECTRA OF THE MAIN STIMULUS B (TRUCK) FOR EACH
AXIS MEASURED FROM THE SEAT SURFACE OF THE TEST BENCH............................. 142
FIGURE 53. UNWEIGHTED POWER SPECTRA OF THE MAIN STIMULUS C (HARVESTER) FOR
EACH AXIS MEASURED FROM THE SEAT SURFACE OF THE TEST BENCH. ................... 143
FIGURE 54. UNWEIGHTED POWER SPECTRA OF THE MAIN STIMULUS D (TRAIN) FOR EACH
AXIS MEASURED FROM THE SEAT SURFACE OF THE TEST BENCH............................. 143
FIGURE 55. UNWEIGHTED POWER SPECTRA OF THE MAIN STIMULUS E (EXCAVATOR) FOR
EACH AXIS MEASURED FROM THE SEAT SURFACE OF THE TEST BENCH. ................... 143
FIGURE 56. THE VISUAL INFORMATION PRESENTED TO PARTICIPANTS TO JUDGE

DISCOMFORT BASED ON THE CONTINUOUS JUDGMENT OF LINE LENGTH (LEFT) AND A
ROTARY CONTROL FOR EVALUATING DISCOMFORT (RIGHT). ................................... 146
FIGURE 57. A TEST SUBJECT EVALUATING DISCOMFORT DURING THE TRIAL. ................... 148
FIGURE 58. SEQUENCE ‘A’ PRESENTED AS THE FREQUENCY WEIGHTED R.M.S. VALUES (M/S2)
USING THE SEAT AND THE BACKREST TRANSLATIONAL AND THE SEAT ROTATIONAL AXES
WITH THE RESPECTIVE MULTIPLYING FACTORS. ..................................................... 151
FIGURE 59. SEQUENCE ‘B’ PRESENTED AS THE FREQUENCY WEIGHTED R.M.S. VALUES (M/S2)
FIGURE 60. SEQUENCE ‘C’ PRESENTED AS THE FREQUENCY WEIGHTED R.M.S. VALUES (M/S2)
FIGURE 61. AN EXAMPLE OF CONTINUOUS JUDGMENT OF VIBRATION (LOWER FIGURE SHOWS

ACCELERATION OF SEAT VERTICAL AXIS, ABOVE THE DISCOMFORT JUDGMENT OF THE
STIMULUS IS SHOWN). ......................................................................................... 153
FIGURE 62. AVERAGED JUDGMENTS FOR EACH STIMULUS FOR ALL SUBJECTS, AND MEN AND
WOMEN SEPARATELY. ......................................................................................... 155
FIGURE 63. CORRELATION BETWEEN MEAN JUDGEMENTS OF ALL SUBJECTS AND

FREQUENCY WEIGHTED VIBRATION MAGNITUDES USING THE R.M.S. METHOD AND
RESPECTIVE MULTIPLYING FACTORS FOR CHOSEN SCENARIOS. .............................. 166
FIGURE 64. DIFFERENCES IN CORRELATION FOR STIMULI TYPE A FOR SCENARIOS 1 (LEFT)
AND 2 (RIGHT) USING AVERAGED JUDGMENTS FROM ALL SUBJECTS. ....................... 172
XXIV
FIGURE 65. DIFFERENCES IN CORRELATION FOR STIMULI TYPE B FOR SCENARIOS 1 (LEFT)
FIGURE 66. DIFFERENCES IN CORRELATION FOR STIMULI TYPE C FOR SCENARIOS 1 (LEFT)
FIGURE 67. DIFFERENCES IN CORRELATION FOR STIMULI TYPE D FOR SCENARIOS 1 (LEFT)
FIGURE 68. DIFFERENCES IN CORRELATION FOR STIMULI TYPE E FOR SCENARIOS 1 (LEFT)
FIGURE 69. CORRELATION OF AVERAGED JUDGMENTS AND STIMULI TYPES USING SCENARIO
1 (SEAT TRANSLATIONAL AXES WITH 1.0 MULTIPLYING FACTORS FOR ALL AXES). ..... 173
FIGURE 70. CORRELATION OF AVERAGED JUDGMENTS AND STIMULI TYPES USING SCENARIO
2 (SEAT HORIZONTAL AXES WITH 1.4 MULTIPLYING FACTORS). ............................... 173
FIGURE 71. CONTOUR MAP OF THE SEAT TRANSLATIONAL AXES MULTIPLYING FACTORS
WHERE THE VERTICAL AXIS IS 1.0. THE ELEVATION (I.E. THE CORRELATION R) IS
INCREASED FOR WARMER COLOURS. .................................................................... 189
FIGURE 72. LEFT FIGURE SHOWS THE CONTOUR MAP SLICED FROM THE SEAT LATERAL
FACTOR AT 1.4 AND 1.8. RIGHT FIGURE SHOWS CURVES SLICED FROM THE SEAT FORE-
AND-AFT FACTOR AT 1.4 AND 2.7. ........................................................................ 189
FIGURE 73. CONTOUR MAP OF THE SEAT TRANSLATIONAL AXES’ MULTIPLYING FACTORS
WHERE THE VERTICAL AXIS IS 1.0. ........................................................................ 190
FIGURE 74. CONTOUR MAP OF SEAT TRANSLATIONAL AXES’ MULTIPLYING FACTORS WHERE
VERTICAL AXIS IS 1.0. .......................................................................................... 191
FIGURE 75. SPEARMAN CORRELATION (R2) FOR EACH TEST SUBJECT FOR SCENARIOS 1, 2
AND 8. ................................................................................................................ 196
FIGURE 76. ASPHALT SURFACES USED IN THE TRIAL. .................................................... 204

FIGURE 77. COBBLESTONE ROAD SURFACES USED IN THE TRIAL.................................... 204
FIGURE 78. OFF-ROAD SURFACE USED IN THE TRIAL. .................................................... 204
FIGURE 79. GRAVEL ROAD SURFACES USED IN THE TRIAL.............................................. 205
FIGURE 80. CONSTRUCTION GRAVEL ROAD SURFACE USED IN THE TRIAL. ...................... 205
FIGURE 81. THE TEST VEHICLE USED IN THE FIELD TRIAL. SEVEN SUBJECTS WAS TESTED
SIMULTANEOUSLY. .............................................................................................. 207
FIGURE 82. SEATING AND MEASUREMENT ARRANGEMENT INCLUDING SEAT ID NUMBERS

(TOP VIEW ). CIRCLES WITH DIAGONAL PATTERNS REPRESENT 6-AXIS MEASUREMENT
LOCATIONS, A CIRCLE WITH A HORIZONTAL PATTERN REPRESENTS THE 12-AXIS
MEASUREMENT LOCATION (DRIVER). .................................................................... 207
FIGURE 83. SENSOR CONFIGURATION FOR EACH SUBJECT (LEFT) AND LOCATION OF THE
BACKREST SENSOR (RIGHT). ................................................................................ 208
FIGURE 84. TITAN (LEFT) AND HERCULES (RIGHT) MEASUREMENT DEVICES USED IN THE
FIELD TRIAL. ....................................................................................................... 208
FIGURE 85. THE DEVELOPED REMOTE DIAL PAD GIVEN FOR EACH SUBJECT TO JUDGE
DISCOMFORT. A TAPE AROUND THE DIAL INDICATED SCALING FROM “NO DISCOMFORT”
TO “HIGH DISCOMFORT”. ...................................................................................... 209
FIGURE 86. OULU COUNTY AREA FROM UP ABOVE AND THE WHOLE TEST ROUTE (YELLOW
LINE) USED IN THE TRIAL. ..................................................................................... 209
XXV
FIGURE 87. A TEST SUBJECT CONDUCTING A JUDGMENT DURING THE MEASUREMENTS. .. 210
FIGURE 88. ILLUSTRATION OF OVERLAPPED VALUES CALCULATED FROM THE VIBRATION
DATA. ................................................................................................................. 212
FIGURE 89. SPEARMAN CONFIDENCE TABLE FOR DETERMINING THE GOODNESS OF A

CORRELATION VALUE COMPARED TO A SAMPLE SIZE. ............................................. 214
FIGURE 90. FREQUENCY WEIGHTED POWER SPECTRAL DENSITY FUNCTION OF ASPHALT

ROAD WITH SPEED HUMPS (LEGS 1 AND 9) AND HIGHWAY ASPHALT (LEG 11) FOR ALL
TRANSLATIONAL AXES FROM THE SEAT (SUBJECT 5 AND GROUP 1). ........................ 217
FIGURE 91. FREQUENCY WEIGHTED POWER SPECTRAL DENSITY FUNCTION OF

COBBLESTONE ROAD (LEGS 3 AND 4) AND GRAVEL ROAD (LEGS 6, 7 AND 10) FOR ALL
TRANSLATIONAL AXES FROM THE SEAT (SUBJECT 5 AND GROUP 1). ........................ 218
FIGURE 92. FREQUENCY WEIGHTED POWER SPECTRAL DENSITY FUNCTION OF OFF-ROAD

(LEG 8) FOR ALL TRANSLATIONAL AXES FROM THE SEAT (SUBJECT 5 AND GROUP 1). 218
FIGURE 93. FREQUENCY SPECTRA (WITHOUT WEIGHTING) FOR TWO DIFFERENT SENSOR
LOCATIONS FOR BACKREST VERTICAL AXIS. BLUE LINE IS FOR THE REFERENCE
LOCATION WHERE THE SENSOR WAS INSTALLED BETWEEN THE SUBJECT AND THE
CUSHION, AND GREEN LINE IS WHERE THE SENSOR WAS INSTALLED OUTSIDE THE
BACKREST, BEHIND THE SUBJECT. ........................................................................ 219
FIGURE 94. AN EXAMPLE OF DELAY BETWEEN THE RAW JUDGMENT DATA AND THE
UNWEIGHTED ACCELERATION FOR THE VERTICAL AXIS FOR SUBJECT 2 AND FOR LEG 3.
......................................................................................................................... 220
FIGURE 95. JUDGMENT STYLES OF GROUP 1 FOR LEG 1 WITH SMOOTH ASPHALT AND SPEED
HUMPS (SUBJECT 6 WAS NOT INCLUDED, BECAUSE OF INVALID DATA). .................... 221
FIGURE 96. DIFFERENT JUDGMENT STYLES OF GROUP 1 FOR LEG 2 WITH SMOOTH ASPHALT
(SUBJECT 6 WAS NOT INCLUDED, BECAUSE OF INVALID DATA)................................. 222
FIGURE 97. LEG 8 JUDGED BY ALL SUBJECTS FROM THE GROUP 1 (SUBJECT 6 WAS NOT
INCLUDED, BECAUSE OF INVALID DATA). ................................................................ 223
FIGURE 98. CORRELATION (SPEARMAN R2) OF DIFFERENT SETUPS TO A STANDARD

SCENARIO 2 (THE SEAT TRANSLATIONAL AXES WITH 1.4 MULTIPLYING FACTORS FOR
THE HORIZONTAL AXES) FOR SUBJECT 3. SETUPS ARE EXPLAINED IN TABLE 56. ...... 226
FIGURE 99. CORRELATION (SPEARMAN R2) OF DIFFERENT SETUPS TO A STANDARD

SCENARIO 2 (THE SEAT TRANSLATIONAL AXES WITH 1.4 MULTIPLYING FACTORS FOR
THE HORIZONTAL AXES) FOR SUBJECT 14 (GROUP 2). SETUPS ARE EXPLAINED IN TABLE
56...................................................................................................................... 227
FIGURE 100. CORRELATION (SPEARMAN R2) FOR ALL SETUPS FOR EACH SUBJECTS USING
STANDARD SCENARIO 2 (SEAT TRANSLATIONAL AXES WITH 1.4 MULTIPLYING FACTORS
FOR HORIZONTAL AXES). ..................................................................................... 229
FIGURE 101. THE JUDGMENT DATA AVERAGED USING SETUP 4B (20 SECOND WINDOW , 10
SECOND OVERLAP AND 2.5 SECOND DELAY) FOR THE WHOLE ROUTE FOR SUBJECT 1.
......................................................................................................................... 230
FIGURE 102. NORMALISED JUDGMENT VALUES OF ALL SUBJECTS FROM GROUP 1 (LEFT) AND
AN AVERAGED JUDGMENT VALUE OF THE GROUP (RIGHT) FOR SETUP 4B (20 SECOND
WINDOW , 10 SECOND OVERLAP AND 2.5 SECOND DELAY). ..................................... 230
FIGURE 103. CLUSTERING OF DATA POINTS USING NEW MULTIPLYING FACTORS OF EACH
GROUP USING SETUP 4A DATA. ............................................................................ 234
XXVI
FIGURE 104. CLUSTERING OF MULTIPLYING FACTORS FOR SEAT HORIZONTAL AXES
(MULTIPLYING FACTOR VERTICAL AXIS IS 1.0) FOR THE THREE GROUPS USING A
CONTOUR MAP. ................................................................................................... 234
XXVII
Abbreviations
OVTV Overall Vibration Total Value
PVTV Point Vibration Total Value
VDV Vibration Dose Value
r.m.s. Root mean square
r.m.q. Root mean quad
r.s.s. Root-sum-of-squares
WBV Whole-body Vibration
LBP Low Back Pain
BS British Standards
ISO International Standards Organization
m/s2 Metres per second squared
PSD Power Spectral Density
x-axis Fore-and-aft axis
y-axis Lateral axis
z-axis Vertical axis
av Overall vibration total value combined using root-sum-of-

squares method
XXVIII
1 Chapter 1 – Introduction
1.1 Background
Vibration exposure can occur at work, commuting between home and work, and in
leisure activities. Any form of transportation will expose humans to some degree of
vibration. Exposure to vibration can cause health and comfort problems. Health
problems are normally back and neck related, presenting as musculoskeletal pain.
Back pain, to which vibration exposure might be a significant contributor (Griffin 1990),
is one of the most common health problems in the world (Deprez, et al 2005, Hostens
2004). Comfort on other hand is related to both physiological and psychological factors,
which can have wider range of effects.
Even though health is the most critical issue in general, only a fraction of people are
exposed to vibration that is severe enough to be identified as the sole cause of long-
term health problems (Rehn, et al 2005b). Most exposed people experience vibration
that may cause discomfort, although it also constitutes a risk factor for low back pain.
Discomfort can show as a general emotional or physical annoyance, such as lowered
ability to concentrate, depending on the context and the emotional state of the human
(Wilder, et al 1994). For example, reading in a train is more difficult because of
vibration (Sundström 2006). As growing numbers of people spend time travelling there
is a need to understand and improve the conditions. Although there is not necessarily a
link between discomfort and health, it is generally assumed that improved comfort is
associated with reduced health risk.
It is difficult to study whole-body vibration health effects directly due to ethical

considerations. It also has been proven difficult to find any pathological proof of back
pain using MRI or other scanning techniques (Videman, et al 2000), thus subjective
opinion has been an important factor in cross-sectional studies of injury prevalence.
Even though studies have shown increased prevalence of back pain when exposed to
vibration, it has been difficult to isolate the effects from other confounding factors
(Palmer, et al 2008). Most of the techniques used to evaluate whole-body vibration
health effects are based on perception and comfort studies (Mansfield 2005), thus the
method for evaluating the health effects is the same as for evaluating comfort.
Because of the harmful effects of vibration, research has been conducted to quantify
methods for evaluating the effects. Currently there are several methods available for
evaluating the effects of vibration and an internationally standardised procedure to
1
measure and calculate values (ISO 2631-1 1997), which has been updated with an
amendment in 2009. Also legislation has been produced to minimise the negative
effects of vibration to work health by forcing all employers within the European Union to
assess the risks arising from vibration exposure (Beazant 2005).
1.2 Problem description

Systematic research relating to the effects of vibration to whole-body vibration dates
back to late 1960’s. During the following decade the basis of the current knowledge
was formed (Griffin, et al 1982). The main focus of the studies were the equivalent
comfort contours, which showed that the human sensitivity to vibration was related to
both amplitude and frequency (Griffin, et al 1982). Thus, it was necessary to weight
(i.e. filter) the measured vibration data before evaluating the effects. Additionally some
work was conducted relating to methods for averaging the vibration signal to a single
value and combining the values of different axes (e.g. vertical, lateral and fore-and-aft)
and locations (seat, backrest and floor) (Parsons and Griffin 1983). It was found that
the vibration of different axes should be combined for a best result. The root mean
square (r.m.s.) method was the first generally accepted method to average vibration
data. Later, also root mean quad (r.m.q) and vibration dose value (VDV) were
proposed as alternative methods if the vibration contained high shocks (Griffin 1986,
Parsons and Griffin 1983). Root sum of squares (r.s.s.) was concluded the best method
to combine the averaged values of each axis. A proposal of the method and guidance,
which is in its current form in ISO 2631-1 (1997), was published in 1986 (Griffin 1986).
The method included frequency weighting curves for all axes (translational and
rotational) and locations (seat, backrest and floor), equations to average and evaluate
the vibration data and multiplying factors to emphasise the axes when combining them.
A year later it became a British Standard (BS 6841) for evaluating the effects of
vibration to human health and comfort.
However, the full method described in ISO 2631-1 (1997) and BS 6841 (1987) has
been rarely used and there is very little information on how accurate the method is for
assessing discomfort in a multi-axis environment. There has been no study, which
would have validated the method and described best practices, using either laboratory
and field environments and field-like (i.e. non-stationary, random and multi-axis) stimuli.
There are related studies (Donati, et al 1983, Fairley 1995, Parsons and Griffin 1983),
but they have been conducted either based on earlier standards, earlier versions of the
2
frequency weightings, without multiplying factors, using shock type stimuli on a fixed
track, using short exposures (< 20 seconds) or using less than twelve axes.
Previous publications have indicated that parts of the standard method might not be
valid. There have been doubts expressed regarding frequency weighting curves
(Maeda, et al 2008, Morioka and Griffin 2006), multiplying factors (Maeda and
Mansfield 2006b) and averaging methods (Maeda 2005). It has been suggested that
the current method does not provide accurate results, which are comparable between
environments (Mansfield and Maeda 2005b).
The equivalent comfort contours (i.e. frequency weightings) have been derived using
single frequency sinusoidal or random narrow-band (i.e. 1/3 octave) vibration. The
stimuli have been stationary (e.g. sinusoidal). In practice humans are exposed to non-
stationary vibration, which is random, and skewed by the frequency content depending
on the dynamic properties of the vehicle. No studies have been found to evaluate the
applicability of the standard frequency weightings for non-stationary random vibration
for predicting discomfort. Additionally most studies have used hard seats, thus the
effect of seat cushion has not been included. This might have an effect on the relative
contribution of different frequencies and axes. The lack of validation also concerns
scaling factors and the averaging method.
The full method provides a single value (i.e. overall vibration total value, OVTV), which
represents the combined effect of all axes to discomfort. The standard does not provide
adequate information on how to compare the results or to understand how the value
relates to discomfort. A table provided in the standard to assess discomfort has
overlapping categories and no references have been provided for the validity or scope
of the table.
The full method includes twelve axes measured from backrest, seat and floor, which
are then calculated and combined as a single value (Figure 1). The translational axes
from the seat have been considered the main axes, while rest of the axes are given as
‘additional’ axes. The OVTV value is significantly affected by the number of selected
axes. There have been only a few studies that have investigated the effects of the
‘additional’ axes simultaneously with the other axes (i.e. multi-axis environment). The
axes have been studied mostly in a single-axis environment. The studies indicate
(Maeda 2004, Wyllie and Griffin 2007) that the 3-axis measurements from the seat
might not be enough for evaluating discomfort from whole-body vibration.
3
Vertical (z)
Lateral (y) PVTV

Seat translational
Fore-and-aft (x)
Seat Roll (rx)

PVTV
Pitch (ry)
Seat rotational
Yaw (rz)
OVTV
Vertical (z)
Backrest
PVTV
Lateral (y)
Backrest translational
Fore-and-aft (x)
Vertical (z)
Floor
PVTV
Lateral (y) Floor translational
Fore-and-aft (x)
r.m.s. method r.s.s. method r.s.s. method
Figure 1. ISO 2631-1 (1997) standard method for combining the frequency weighted vibration data
from each of the twelve axes to a single “overall vibration total value”. The r.m.q and VDV methods
can be used instead of the r.m.s. method for averaging the vibration data.
The full method requires equipment that has three separate sensor locations: two
sensors for recording translational axes, and one sensor for recording both
translational and rotational axes. Sensors measuring the translational axes are
common, but rotational acceleration is more difficult to measure and few products are
available, which are expensive and not available off-the-shelf. Because of the lack of
published field measurements there is little information on the usability of measuring all
twelve axes in the field, thus there is no guidance on how many axes are practically
necessary for evaluating the discomfort from vibration. It would encourage more
measurements if fewer axes were needed.
1.3 Objectives and hypotheses

The objective of the research was to validate and, if necessary, improve the full method
of the ISO 2631-1 standard for evaluating discomfort from whole-body vibration. Even
though it was assumed, that the method predicts discomfort at least adequately, there
was a suspicion that parts of the method (e.g. frequency weightings) will not provide
optimal results, thus the method can be further improved. The possible improvements
4
were limited to such changes that would allow previous measurements and current
equipment to be used. This restricts the optimisation to:
• Either frequency weightings are used or they are not (i.e. no new frequency
weightings are produced);
• Either r.m.s. or VDV (or r.m.q) methods are used (no new averaging methods
are suggested);
• No new measurement locations are introduced.
However, there are parts of the method, which can be changed more freely:
• New multiplying factors can be derived for all axes;
• Combinations of axes (i.e. locations and directions) to produce OVTV can be

selected freely.
Further hypotheses, based on the previous studies (see Chapter 2), were:
• Applying frequency weighting will more likely have a small effect on discomfort
predictions as the vibration frequency range found in the field is heavily skewed
at low frequencies (below 5 Hz), which is also emphasised by the weighting
curves (see Section 2.2.1);
• Higher multiplying factors should be used for the seat horizontal axes,
especially if only seat translational axes are used to evaluate discomfort
(Maeda and Mansfield 2006b);
• Emphasis of shocks by using a VDV or r.m.q. method might provide better

correlation to discomfort than the r.m.s. method if the vibration has high shocks
(Boileau, et al 1989);
• Including more axes to the evaluation will improve the correlation to subjective
judgment (Parsons and Griffin 1983);
o However, most likely the difference will be small compared to using just
seat translational axes with correct emphasis (i.e. multiplying factors).
1.4 Thesis structure

A literature review was conducted for providing a comprehensive background on the
evolution and history of the standard method, and to summarise the current knowledge
5
and guidance on the effects from whole-body vibration exposure (Chapter 2). Based on
the review general methods to conduct the thesis were chosen (Chapter 3).
Technical development of measurement equipment was considered necessary for

measuring all twelve axes in the field. The equipment was validated in two experiments
using a multi-axis test bench and compared to commercial equipment. The equipment
was developed to fulfil the requirements for commercial equipment according to the
relevant standard (Chapter 4).
Field measurements of the twelve axes were conducted to evaluate the relative
contribution of the axes to OVTV by analysing the effects of frequency weighting and
multiplying factors. Several typical work environments were measured and results from
previous studies were used to determine the axes, which have a practical effect to
discomfort (Chapter 5).
A laboratory trial was conducted for validating the standard method in a multi-axis
environment by acquiring both vibration and discomfort judgments. Detailed analysis of
combinations of axes, frequency weighting curves, multiplying factors and averaging
methods were conducted to find a procedure for best possible correlation to discomfort.
Additionally the data was used to optimise the standard method based on the allowed
changes (Chapters 6 to 8).
A field trial was conducted for validating the findings from the laboratory trial and to
suggest best practices for predicting discomfort from whole-body vibration. Similar
methods to the laboratory trial were used; however several subjects were measured
simultaneously (Chapter 9).
1.5 Scope
This thesis was limited to a seated posture of a passenger using a seat with a backrest
support and a rigid body frame (i.e. typical car seat used in a passenger car). Each part
of the thesis had specific limitations, but in general this thesis regarded only the effects
of vibration to discomfort, thus other sources (e.g. noise) were neglected. The aim was
not to develop a completely new method or model to predict discomfort, but to
determine the limitations and to improve the current standard method as much as
possible. For this reason the analysis and results were limited to the inherent problems
relating to the theory and assumptions behind the standard method.
6
Based on the methods and environments the correlation results will not necessarily
imply direct causality between subjective comfort and vibration exposure, but only that
the variables have correlation.
1.6 Original contribution to knowledge

The following list summarises the main contributions of the thesis (contributions are
presented in detail in Chapter 12):
1. Development of an affordable and practical 12-axis sensor system for field

measurements;
2. Analysis of the practical importance of the twelve axes in the field based on the
previous studies and new field measurements;
3. Laboratory validation of the axes and correlation between discomfort and

vibration exposure of the standard method;
4. Analysis of the standard method by examining the role of frequency weighting,

averaging method and multiplying factors;
5. Optimisation of new parameters to improve the correlation to discomfort;
6. Validating and testing the new proposed parameters and the current standard
method in field;
7. Guidance on using the standard method with and without the optimised
parameters in practice.
Additionally minor contributions of the thesis are:
1. Testing and validating cross-modal matching method of continuous judgment

for acquiring discomfort judgments from subjects;
2. Detailed analysis of the current standard and its guidance;
3. Proposal of a method to optimise the new parameters for improving the

standard method;
4. Development of equipment to allow judgment data to be gathered in field

simultaneously with vibration;
5. Information on how to average discomfort judgments for best results;
6. Gender differences in discomfort judgments;
7. Guidance on how to conduct discomfort evaluation in practice.
7
1.7 Publications
A number of publications have been published based on the results of the thesis (Table
1). In total eight conference papers and two journal papers (in review) have been
submitted. An additional journal paper regarding the field trial (Chapter 9) is planned to
be submitted later.
Table 1. Publications based on the results of the thesis.
Topic of the paper Venue Year
Requirements for validating the standardized Japan conference on human response to

2006
whole body vibration comfort evaluation method vibration
Improving 12-axis measurement equipment for

Inter-noise 2006
whole-body vibration comfort evaluation
nd
Relative existence of 12-axis acceleration in 42 UK Conference on Human Response to
2007
different directions in mobile machinery work Vibration
rd
Optimising the standardised method to evaluate 43 UK Conference on Human Response to
2008
discomfort from multi-axis whole body vibration Vibration
nd
2 International Conference on Human
Measuring and evaluating discomfort from whole-
Vibration Exposure, measurement and tests, 2009
body vibration in practice
Boden, Sweden
th
Journal of Industrial Health and 4
Relative contribution of translational and rotational
International conference on whole body 2009
vibration to discomfort
vibration injuries
th
The relative contribution of twelve axes of Journal of Industrial Health and 4
vibration in field measurements for analysis International conference on whole body 2009
according to ISO 2631-1 vibration injuries
th
Evaluation of discomfort in the field according to 44 UK Conference on Human Response to
2009
ISO 2631-1 standard Vibration
8
2 Chapter 2 - Literature review
2.1 Introduction and aims of the review
Vibration exposure and its effects lie at the crossroads of many research disciplines:
engineering, ergonomics, psychology, clinical medicine, and physiology. All disciplines
have a similar goal of optimising the environment for the exposed persons, but different
methods of achieving it. Engineering aims to minimise vibration exposure using
technical solutions, while ergonomists try to optimise the environment using a human-
centred approach. Psychology, clinical medicine and physiology concentrate on the
human body and mind, and understanding how vibration causes stress. To find novel
solutions, that will significantly minimise the negative effects from vibration, requires
combination of knowledge across the different disciplines.
There are literally thousands of publications relating to whole-body vibration. The

effects of whole-body vibration on a human body have been a subject of research at
least since the early 20th century (Griffin 1990, Mansfield 2005, Oborne 1977). The
effects are complex, non-linear and depend at least on the vibration amplitude,
direction, frequency, duration and to which part of the body it is directed. A major part
of the studies has been about measuring and analysing the vibration exposure levels
concerning health for various work environments. However, not many have given
consideration to the improvement of the methods currently used for evaluating and
calculating the effects of vibration. There is a consensus that vibration exposure
causes problems, but the mechanics of the causes are still not fully understood.
Based on the research conducted in 1970’s and 1980’s an international standard ISO
2631-1 (1997) was created to guide measurements and evaluation of whole-body
vibration exposure relating to health, discomfort, perception and motion sickness. The
standard is similar to the more concise British Standard BS 6841, which was
introduced ten years earlier in 1987. Both of them use the same principal method for
assessing discomfort, where 1) measured vibration data from selected locations and
axes are frequency weighted (i.e. filtered), 2) averaged to a single value, 3) each
averaged value is multiplied by a factor, 4) the multiplied averaged values for each axis
and location are combined to form an overall vibration total value, which is 5) compared
to a table or other values to assess discomfort from the measured environment.
However there are still doubts and concerns expressed about the methods used for
assessment (Griffin 2007). Based on the standard, legislation has been introduced in
9
Europe where there is a requirement to minimise the effects from vibration exposure
(EC 2002).
Standards and legislation have focussed on managing and preventing health problems,
most particularly to low back pain (LBP). Health effects have had the highest priority in
research, however as most of our knowledge of health effects (and methods to
evaluate it) come from perception and comfort studies, there is a clear link between
health effects and discomfort judgment of a subject.
The purpose of the literature review is to summarise current knowledge on the effects
of whole-body vibration to a seated person (Section 2.2), the methods for
understanding and evaluating discomfort (Sections 2.3 and 2.4), technical
requirements for measuring it (Section 2.5) and guidelines that can be used to assess it
(Section 2.6). The review is concentrated on discomfort effects, but also other effects
are reviewed because of close links between them. Even though there has been a
large number of studies in this area, a relatively small number of publications can be
considered important. The review will go through the literature regarding discomfort in
the form of scientific publications, reports, standards, guidelines and legislation. Based
on the review, the reader can understand more clearly the problems of vibration
exposure and how health and discomfort are related to it.
2.2 Effects from vibration exposure

It is important to understand the effects of whole-body vibration, because millions of
people expose themselves to it every day while working and commuting (Hostens
2004, Mansfield 2005, Palmer, et al 2003). It is almost impossible for any person to
avoid vibration exposure living in a modern society. Just in the UK alone there are
about 8.5 million people (estimated) who are exposed to whole-body vibration at work
every day (Mansfield 2005). Vibration also affects passenger’s ability to carry out tasks,
such as eating, reading and writing (Oborne 1977). Under more extreme conditions
vibration will have psychological and physiological effects, such as motion sickness,
headaches, and even chronic health effects. About 1.3 million people in UK are
exposed to vibration exceeding the action value of the Control of Vibration at Work
Regulations (Brereton and Nelson 2005). Direct medical costs related to low back pain,
which can be caused by vibration exposure, is in billions of dollars (Hostens 2004).
Whole-body vibration is defined as a motion transmitted to the human body as a whole

through supporting surfaces, as opposed to vibration directed more locally, such as
hand-arm vibration. Term “whole-body vibration” applies to standing, seated and
10
recumbent persons (ISO 2631-1 1997). Because sitting is the most common posture in
which a human body interacts with vibration, it is the most researched and understood
(ISO 2631-1 1997). Driving and operating a machine and travelling are most frequently
done in a seated posture. Other postures are rarer in work environments and have
different physical effects to the human body (Mansfield 2005). Other postures are also
more cumbersome to assess. For example a standing person tends to use legs to
attenuate vibration. A recumbent person is on the other hand is normally associated
with situations like patient transport or night-time travel, not generally with work. Both of
these postures also have less association with low back pain than seated posture
(Magnusson and Pope 1998).
Comfort and health effects relate in many ways. The methods used for health
evaluation are mainly based on perception and comfort studies (Mansfield 2005). It is
very difficult to even study health effects directly, because it is unethical to cause health
problems purposely, and no personal long term vibration exposure history is available.
It also has been proven difficult to find any visual proof of back pain using MRI or other
scanning techniques (Videman, et al 2000), thus subjective opinion has been an
important factor in published studies.
2.2.1 Discomfort
Discomfort is a highly subjective and personal opinion which varies between individuals
(Griffin and Whitham 1977, Rybiak 2003), although some consistencies can be found
amongst humans (Griffin, et al 1982, Parsons and Griffin 1980). In whole-body
vibration research the interest has been to link the discomfort feeling to the vibration
characteristics, such as frequency, amplitude, direction and duration, and to human
characteristics, such as height, weight, age and gender. Human reaction to vibration is
dependent on three factors: 1) characteristics of the vibration, 2) characteristics of the
human and 3) characteristics of the environment. A seated person on a typical seat will
usually perceive vibration from three locations (seat surface, backrest and floor). In a
normal situation the vibration to a human is multi-directional, meaning that it is a
combination of three translational axes (vertical, lateral and fore-and-aft) and three
rotational axes (pitch, roll and yaw) (Figure 2). Back and floor vibrations can be
analysed using translational axes, because in a rigid-seat environment the rotational
axes can be assumed to be the same for floor and backrest as for the seat surface.
11
Figure 2. The twelve axes and locations used to determine vibration discomfort for seated
subjects (Griffin, et al 1982).
In the context of work, discomfort can be related to fatigue, lowered concentration and
work performance, and indirectly to work motivation and happiness (Wertheim 1998),
which can increase workload and reduce performance (Newell and Mansfield 2006).
Vibration can also cause drowsiness and irritation (Wertheim 1998). These factors are
becoming more and more important in work environments. A worker-friendly workplace
can also show benefits in productivity (Hoy, et al 2005). Although there is not
necessarily a link between discomfort and health, it is generally assumed that improved
comfort is associated with reduced health risk. Even though vibration is rarely the only
contributor, it is possible to improve overall comfort by reducing it (Hostens 2004,
Mansfield 2005).
Forming an opinion of vibration starts immediately when a human perceives it. There
are different sensory systems in the human body that can sense the vibration. There is
no single organ which the human uses to perceive the vibration or movement. The
systems can be divided into visual, vestibular, somatic and auditory systems (Mansfield
2005). The feeling of discomfort is formed by the combination of these systems. Each
of the functions perceive vibration in a different way. The eyes can clearly see high-
displacement movement, which normally occurs in relatively low frequencies. The
perception is based on relative changes in position with a reference object (i.e.
movement of a building cannot be normally detected with other organs). The vestibular
system of the inner ear senses linear and rotational acceleration. When a head is
exposed to acceleration or changes orientation to gravity, the vestibular system detects
the movement. The somatic system includes many elements, which sense the vibration
through motion of joints and muscles and receptors in skin and abdomen. Different
receptors underneath the skin can detect vibration up to 500 Hz. The auditory system
12
on the other hand can indirectly detect vibration when motion produces change in air
pressure (above 20 Hz). Depending on the vibration characteristics and environment
where a human is exposed to it, the combination of the sensory systems provides
information to the brain to form a cognitive model of the motion (Mansfield 2005).
Additionally human comfort depends on both static and dynamic factors. Static comfort
relates more to ergonomics, while dynamic comfort relates to the vibration
characteristics.
Discomfort is a complicated term to define. Depending on the context and purpose it

can be interpreted in many ways. Regarding work, the threshold for no (or acceptable)
discomfort can be defined as “a state where vibration does not hinder work
performance” or “an environment where vibration does not cause any action to change
it” (Mansfield 2005). For passengers not conducting any obligatory tasks, the
discomfort can relate to general annoyance, inability to fall asleep, and difficulties for
reading and writing (Oborne 1977). The effects will change depending on how vibration
characteristics relate to other sources, such as noise, and from other factors such as
cost of the trip, expectations, tiredness, and so on. In the context of moving machines
vibration can be regarded as always causing discomfort, and decrease of vibration will
always decrease discomfort (it should be noted that in some special cases passengers
expect vibration and will feel ‘uncomfortable’ without it e.g. helicopter transportation). In
this study the term “discomfort” is used in the context of “ride comfort” of passengers.
In this case the change in discomfort can relate to many things, such as improved work
performance (i.e. reading or writing), decrease of annoyance, improved concentration
or even improved happiness.
If one searches terms “whole-body vibration” and “comfort“ or “discomfort” using

databases such as EngineeringVillage2, InformaWorld, SCOPUS and SpringerLink,
which incorporate most of the engineering journals regarding whole-body vibration, a
relatively low number of publications is found (Table 2). EngineeringVillage2 shows a
total of 236 hits where some studies are duplicate from other keyword searches and
some had no relevance to the topic of the thesis. InformaWorld shows 106
publications, which most of them duplicates to EngineeringVillage2. SCOPUS shows
90 and SpringerLink 26, which all of them duplicates from previous database searches.
Less than thirty of the records were published in the 1960’s and 1970’s, 23 records in
the 1980’s, 18 in the 1990’s and 25 between 2000 and 2009. Search term “rotational”
or “rotation” provides only six relevant studies which were all made over 20 years ago.
For an example term “whole-body vibration” alone produced over 1000 publications
13
from the databases. Comfort research has not been popular in recent decades,
possibly because the required knowledge to provide a contribution and lack of
equipment (i.e. vibration test bench and 12-axis measurement device).
Table 2. Number of publications found from database referring to whole-body vibration and
discomfort. EngineeringVillage2 database includes Compendex, Inspec, NTIS, GEOBASE,
Referex and other engineering databases. Informaworld is site hosting journals and ebooks
published by Taylor & Francis, Routledge, Psychology Press and Informa Healthcare. SCOPUS
is the largest abstract and citation database of research literature and quality web sources.
SpringerLink is an integrated full-text database for journals and books published by Springer.
Search was made using keywords, abstracts and titles.
The number of studies in

the order of search words for each
Database Search words used for all databases database
EngineeringVillage2 1. “whole body vibration” AND “discomfort” 90+57+34+55
Informaworld 2. “whole body vibration AND “comfort” 41+42+7+16

3. “whole body vibration” AND “ride
SCOPUS 32+24+3+31
comfort”
SpringerLink 4. “whole body vibration” AND (“rotational” 1+8+2+15

OR “rotation”)
Perception, motion sickness and discomfort are linked together. Motion sickness
though is more of a conflict between the sensory systems than purely the cause of the
vibration (Mansfield 2005). Perception can be thought as the ultimate lower threshold
for discomfort to occur, because it defines the lowest level of vibration that a human
can feel.
2.2.1.1 Studies and results behind the ISO 2631-1 (1997) method
The method described in ISO 2631-1 (1997) has been derived from the same studies
as BS 6841 standard, which was introduced in 1987 (British Standards Institution
1987). Thus all studies before 1987 can be considered to relate to the standard
method.
The experiments conducted in 1960’s by Miwa (Miwa 1969, Miwa 1967, Miwa and
Yonekawa 1974) and 1970’s and early 1980’s by Griffin, Parsons, Whitham and
Corbridge (Corbridge and Griffin 1986, Griffin et al. 1982, Parsons and Griffin 1982,
Parsons et al. 1982, Parsons and Griffin 1983) show the first systematic results on
discomfort and perception thresholds to whole-body vibration for different axes. The
studies focussed on axes that existed in a seated posture. It was the purpose of the
studies to find out the effects of vibration on discomfort regarding frequency, amplitude,
14
direction and location (Griffin, et al 1982). The main results of the studies were the
equivalent comfort contours for each axis in relation to frequencies, directions and
locations. Although there were other studies regarding perception and sensation as
well (Oborne 1978a, Reiher and Meister 1932, Shoenberger and Harris 1971), the
results from the aforementioned studies are the basis of current ISO 2631-1 (1997) and
BS 6841 (1987) standards (Maeda 2005).
Different methods were used for finding out the comfort contours. Some used rating
scales (Fothergill 1972, Jones and Saunders 1974, Oborne and Clarke 1974) while
others used reference stimuli (Griffin, et al 1982, Parsons and Griffin 1982, Parsons, et
al 1982). Almost all of the studies concentrated on one axis (e.g. vertical) at a time. All
studies were made in a laboratory using sinusoidal or other artificial stimuli.
The axis in the first studies for evaluating the sensation contours was vertical. Miwa
(1967) concluded that sensitivity to vibration was higher at lower frequencies and it
decreased for higher frequencies for the same amplitude. This was also confirmed by
Griffin et al. (1982) and other studies. Based on these studies it was shown that for
vertical direction the human perception was most sensitive at around 5 Hz (Griffin, et al
1982). Although results were not exactly the same they showed similar tendencies.
Figure 3 shows the comparisons of experiments and results for vertical direction made
by Miwa (1967), Jones and Saunders (1972), Shoenberger and Harris (1971) and
Griffin (1976). Figure 4 shows the median contours found by Griffin et al. (1982). The
results are also consistent with the previous studies. No significant differences were
found between men and women (Griffin, et al 1982).
15
Figure 3. Median equivalent comfort contours of different studies for vertical axis at seat for a
seated person compared between previous studies (- - -) and the reference study (---) (Griffin, et
al 1982).
Figure 4. Median, 25th and 75th percentile comfort contours for vertical seat vibration (Griffin,
Whitham and Parsons 1982).
16
The human body seemed to be most perceptive to horizontal vibration at very low
frequencies (1 to 2 Hertz) and perception became more insensitive at higher
frequencies (Griffin, et al 1982) (Figure 5). The studies did not find clear differences
between lateral and fore-and-aft directions (without backrest). Results were compared
also with the previous studies and similarities were found (Figure 6).
Figure 5. Median, 25th and 75th percentile comfort contours for fore-and-aft (left) and lateral
(right) seat vibration (Griffin, et al 1982).
Figure 6. Median equivalent comfort contours for fore-and-aft vibration, compared between
Griffin et al (1982) and Miwa (- - -) (Miwa 1967).
17
Miwa and Yonekawa (1974) published equal sensation curves for all translational axes
using the vibration greatness method (Figure 7). There seemed to be a difference
between sensations if sinusoidal or random vibrations were used. Random vibrations
were felt more severe at higher frequencies. In contrast, sensation was more severe
using sinusoidal stimuli at lower frequencies. Similar curves using sinusoidal stimuli
(Figure 7: left) were obtained for both seated and standing postures.
Figure 7. Miwa’s studies on seat translational axes using vibration greatness method for
sinusoidal stimuli (left) and random stimuli (right) (Miwa and Yonekawa 1974).
In normal transport conditions the human sits on a seat using a backrest and feet
resting on a floor. In this case the vibration will also be introduced through these
locations as well and will most likely affect the comfort. Parsons et al. (1982) made a
comprehensive study of the effects of seat and floor vibrations to subjective comfort.
The conclusion was that the fore-and-aft direction had the most significant effect for
back vibration (Figure 8: left). Vertical and lateral axes had similar effect, which was 5
to 10 times less than for fore-and-aft axis. All three translational axes from the floor had
similar effect (Figure 8: right). At lower frequencies feet were less sensitive to vibration
as felt from the seat surface, but were more sensitive to higher frequencies. The
threshold on the backrest and floor also increased with higher frequencies as seen with
the horizontal directions on the seat.
18
Figure 8. Comfort contours for three translational axes of back (left) and foot (right) vibration
(Parsons, et al 1982).
In addition to translational motion, rotational motion also appears in many work

environments (e.g. forestry harvester). If the centre of rotation is far from the human
then it is perceived as translational motion (Parsons and Griffin 1982), but if the centre
of rotation is close to the human it has its own effects and may contribute to discomfort.
Rotational acceleration (rad/s2) can be separated in to three independent axes, which
are roll, pitch and yaw. Not many studies investigated the effects of pure rotational
accelerations separately in laboratory environment. Based on the studies that did
(Parsons and Griffin 1978a, Parsons, et al 1978, Parsons and Griffin 1982) the
perception of the acceleration was decreased for higher frequencies (Figure 9).
Figure 9. Median, 25th and 75th percentile comfort contours for roll (left), pitch (middle) and yaw
(right) directions (Parsons and Griffin 1982)
The main conclusions from the early perception studies were: 1a) the horizontal axes
(lateral and fore-and-aft) were considered equally uncomfortable and different to
vertical axis and 1b) the rotational axes from the seat, backrest and floor translational
axes had also an effect to discomfort, 2) there were differences between subjects
depending on their physical characteristics 3) sensitivity to vibration seemed to relate to
19
frequency, where higher frequencies (> 10 Hz) were more likely to be perceived less
uncomfortable than lower frequencies (< 10 Hz). Vertical axis showed more sensitivity
to higher frequencies than other axes. It was also concluded in the studies that
vibration amplitude did not affect the comfort contours significantly and there were no
differences between male and female.
It was found early on that human perception is non-linear depending on the frequency
characteristics and amplitude of the vibration exposure. Oborne (1977) found that at
lower intensity levels the higher frequency components (8-16 Hz) had more effect on
comfort than lower components (0-4 Hz). However, when intensity increased to about
1.4 m/s2 then lower frequency components became dominant factors for discomfort.
This was a logical result as high intensity vibration at low frequencies induces motion
sickness and hinders motor tasks such as reading.
The majority of the reviewed studies were concerned with human perception to
different frequencies and amplitudes. Only few studies validating different averaging
methods were found (Fothergill and Griffin 1977, Griffin and Whitham 1976, Parsons
and Griffin 1983). The studies found that sinusoidal frequencies and random vibration
can be effectively predicted using frequency weighting and r.m.s. (root mean square)
averaging. Additionally the r.m.q. (root mean quad) method was suggested over the
r.m.s. method if shocks or impulsive vibration was experienced (Griffin and Whitham
1980a, Griffin and Whitham 1980b). Griffin and Whitham (1977) showed that
discomfort from multiple directions can be summed using r.s.s. (root-sum-of-squares)
method, even though it was noted to be unreliable at least for some combinations of
rotational and translational vibration (Parsons and Griffin 1978b). However the study
case was limited. Parsons and Griffin (1983) applied the knowledge from the early
laboratory studies to analyse correlation to discomfort by effects of different weightings
and methods for combining axes. The conclusions showed that combining the axes will
have better results than using the worst axis or worst frequency. The best correlation
was found when using frequency weighted r.m.s. method and r.s.s. for combining the
axes. However, little differences were found between the r.m.s. and r.m.q. methods.
The effect of duration of exposure on discomfort was also evaluated in the 1960’s and
1970’s by several authors (Griffin and Whitham 1976, Griffin and Whitham 1980b,
Miwa and Yonekawa 1974). There was no clear evidence that relative discomfort
changed when duration was increased (Griffin and Whitham 1976). However, the same
authors later found that longer stimuli (up to 2 minutes) was more uncomfortable than
shorter stimuli (Griffin and Whitham 1980b). Miwa and Yonekawa (1974) found a time
20
dependency between vibration greatness level and exposure time. The effective range
of exposure time was from 10 to 500 minutes. Griffin and Whitham (1980b) also found
a fourth power relationship between duration and vibration magnitude to discomfort
(doubling the amplitude required a 16-fold reduction in duration to maintain
equivalence). The study used durations up to 32 seconds. However, no consistency
and consensus regarding the time dependency was concluded in these studies (other
than showing that a dependency most likely exists).
No publication regarding the validation of multiplying factors for combining several axes
of vibration were found, which could be used to understand the factors presented in the
standard (see Section 2.6 for explanation of the standards). However, the results from
the aforementioned studies showed that the same vibration stimulus produced a
different level of response depending on the axis and location. Thus, combining the
responses of each axis requires a compensation factor to equalise the effect. The
author assumes that this was the basis for the multiplying factors, even though this has
not been stated in any known publication..
In addition to perception and comfort studies, biomechanical studies were undertaken

to provide understanding of human behaviour under vibration (Mansfield 2005).
Findings suggested that the human body has a main resonance (i.e. Eigen frequency)
in the vertical axis around at 4-5 Hz. This was in the same area as perception and
comfort studies concluded for being the most sensitive. The body’s resonance
behaviour in a seated posture in vertical axis was linked to the lumbar spine (Sandover
1988). The results for the horizontal axes (lateral and fore-and-aft) showed resonance
frequencies in the area of 2-4 Hz.
Only one field study was found, which used all twelve axes and subjective judgments,
for proposing and validating a new method based on the earlier studies (Parsons and
Griffin 1983). The study used several cars in test track conditions to measure the
twelve axes of vibration and a judgment from the subjects for twelve different road
types. Most surfaces did not include high shocks, although a wide variety of different
vibration characteristics was achieved. The subjects were driven individually for all
tracks and cars. The subjects rated discomfort using a paper with a line labelled “little
discomfort” and “much discomfort” on each end and marked a cross on the line to
indicate discomfort, which was then converted to a number by measuring the distance
of the cross from the start of the line. The study used weighting curves, which were
created based on a laboratory study with the same subjects and tried several
21
averaging methods (r.m.s., r.m.q. and maximum frequency level) and methods for
combining the axes (r.s.s., r.s.q. (root-sum-of-quads) and worst axis).
The results showed that even though there was high variability between individual
subjects, the trend was similar: 1) r.m.s. and r.m.q. methods were better for averaging
vibration than using the maximum frequency level (no significant differences were
found between the r.m.s. and r.m.q. methods), 2) r.s.s. and r.s.q. methods showed little
differences, but were better than using the worst axis. The conclusion from the study
was that in general the best overall method was to use the r.m.s. method to average
individual axes and to combine them using the r.s.s. method. Additionally it was found
that the seat vertical axis was the most dominant cause of discomfort. From the other
axes fore-and-aft axes from the seat and backrest, and vertical axis from the floor had
importance. Least important axes were the rotational axes and lateral axes from the
floor. Despite high confidence of the results, the study was limited. The publication
cannot be used direcly to assess the current standard method as it used frequency
weightings, which were pre-standard versions and had no multiplying factors for the
axes.
2.2.1.2 Laboratory studies and results validating the standard method
Equivalent comfort contours (i.e. weightings) in ISO 2631-1 (1997) (see Section 2.6.1)
are based on an assumption that magnitude does not affect the relative weighting of
frequencies. More recent studies have found that comfort contours are magnitude
dependent (Ahn and Griffin 2008, Mansfield and Maeda 2005b, Morioka and Griffin
2006). Morioka and Griffin (2006) show consistency with the standard weightings,
although at higher frequencies they show higher discomfort (Figure 10). More recent
studies have found that the rate of increase in discomfort varies with the frequency and
direction of vibration. For example, the rate of increase tends to be greater with low-
frequency vibration than with high-frequency. It has been found that equivalent comfort
contours depend on vibration magnitude, thus becoming less flat with increasing
magnitude (Griffin 2007).
22
Figure 10. Comparison of BS 6841 standard weightings to newer findings. Effect of vibration
magnitude on frequency weightings (inverted equivalent comfort contours normalised at 2Hz for
fore-and-aft and lateral vibration and normalised at 5Hz for vertical vibration): (a) fore-and-aft,
(b) lateral, (c) vertical. A sensation magnitude of 100 is equivalent to the discomfort produced by
2 2
1.0 m/s rms (fore-and-aft and lateral vibration) or 0.5 m/s rms (vertical vibration) at 20 Hz
(Morioka and Griffin 2006).
23
It has been found that the horizontal frequency weighting (W d) of the standard seems to
underestimate the vibration effect compared to vertical direction (Maeda and Mansfield
2006b). The results from Maeda and Mansfield (2006b) showed that subjective
responses had a same category value for all axes exposed to flat band stimuli from 1 to
20 Hz. When frequency weighting was applied the fore-and-aft and lateral axes
showed smaller r.m.s. values, thus underestimating the response compared to the
vertical axis. The authors concluded that a factor of 2.47 for the horizontal axes would
equalise the difference. The results were in agreement with previous studies (Griefahn
and Broede 1999, Mansfield and Maeda 2005a).
There have been also other studies for modelling human body, investigating standing
and recumbent subjects, investigating combined effects of noise and vibration, and so
on, but they can be regarded as out of the scope of the thesis.
2.2.1.3 Field studies and results validating the standard method
Most field studies, after the introduction of the standard, have only measured seat
translational axes and not acquired judgments from the subjects, thus only partial
validation can be made based on the results.
Alem et al. (2004) recommended not using the r.m.s. method at all for evaluating
shocks or VDV for evaluating repeated shocks. Similar findings were concluded in
other earlier studies as well (Griffin and Whitham 1980a, Kjellberg and Wikström 1985),
which indicated that higher exponents than 2 would be better for evaluating shocks.
Boileau et al. (1989) concluded that the r.m.s. method would underestimate the
vibration when including shocks of high crest factors. These results were derived from
measurements of forestry machines (skidders).
Wikström et al. (1991) showed that for most environments with shocks the r.m.s.
method showed high correlation and no significant improvement was found using
higher exponents (i.e. r.m.q.). The conclusion was that it was more important to
consider to whole event instead of just evaluating the peak. The correlation for the
r.m.s. method improved with longer duration of the window size (i.e. number of
samples included in the calculation). Dose methods (i.e. VDV) were concluded for
providing a good correlation as well, but if the compared events have similar durations,
there is no advantage using the dose method. The study concluded also that worst
axes showed higher correlation than the r.s.s. of the axes. However, the summation
and weighting of axes was based on equal multiplying factors. Wikström et al. (1991)
also noted that the proposed standard weighting curve gave worse correlation than
24
using a flat weighting. They tested higher cut-off frequency (from 2 Hz to 4 Hz) and
concluded improved correlations. Even though they used only seat axes, they found no
justification for including all twelve axes.
Els (2005) used a military vehicle to expose subjects to high vibration magnitudes
having also high emphasis on roll motions. He showed that the vertical axis had
highest correlation using any method (e.g. weighted and unweighted r.m.s., VDV, VDI
2057 and Averaged Absorbed Power). Pitch and roll axes were experienced as
translational axes, in this case vertical. The conclusion also was that there was no
difference between the r.m.s. and VDV methods. The unweighted r.m.s. value showed
similar results as the weighted r.m.s. value, which was assumed to be linked to the
suspension system, which attenuates higher frequencies.
Mehta and Tewari (2000) reviewed several papers relating to seating discomfort for
tractor operators. They concluded that no exclusive model or results were found to
support any single method. Fairley (1995) concluded that the best procedure for
predicting discomfort from vibration was that recommended by earlier ISO 2631
standard (1978). In this case the r.s.s. of the frequency weighted r.m.s. values showed
best correlation. Monsees et al. (1988) found that the r.m.q. method had no advantage
over r.m.s. method for predicting discomfort of tractor vibrations.
The results from the studies are summarised in Table 3. In overall the r.m.s. and r.m.q.
methods were found to be the best for averaging the axes. Additionally VDV was
suggested, but it is similar to the r.m.q. method. The r.s.s. method for combining the
axes was found to be much better than just using the most severe axis, even though
some conflicting results were found.
25
Table 3. Summary of field studies regarding suggested methods for manipulating vibration
measurements to predict discomfort (including studies before and after the standard).
Author(s) Study description Metrics tested Conclusions

The most severe axis, the It is recommended to use the
Griffin and concept of masking and the r.s.s. method to combine axes for
To assess discomfort from
Whitham r.m.s. of the equivalent levels practical purposes, but other
dual-axis whole-body vibration
(1977) of both axes methods should be further
investigated
The most severe component,
the r.m.s. of the equivalent The most severe component
To study the effect of position
levels of the single-axis method gives the closest results,
Parsons and of the axis of rotation to
components and the but no significant differences
Griffin (1978b) discomfort for roll and pitch
arithmetic sum of the were found compared to the
axes
equivalent levels of the r.m.s. method
single-axis components
Griffin and To study discomfort produced The r.m.q. method is suggested if
The r.m.s. and r.m.q.
Whitham by impulsive whole-body vibration contains shocks or
methods
(1980a) vibration impulsive vibration
Griffin and To study time-depency of The r.m.q. method is suggested if
The r.m.s. and r.m.q.
Whitham whole-body vibration vibration contains shocks or
methods
(1980b) discomfort impulsive vibration
r.m.s. and r.m.q. methods,
and maximum frequency for r.m.s. to average each axis and
Parsons and Field study, which recorded 12
averaging axes and r.s.s., r.s.s. to combine axes were
Griffin (1983) axes and subjects’ judgments
r.s.q. and worst axis to found the best combination
combine axes
To study relationship between
The r.m.q. method had no
Monsees et al. subjective assessment and The r.m.s. and r.m.q.
advantage over the r.m.s.
(1988) objective measurements of methods
method in tractor vibration
tractor ride
The r.m.s. method provides
To study and to compare the
Boileau et al. The r.m.s, r.m.q and VDV systematically lower values with
results of the r.m.q. method to
(1989) methods vibration including shocks
ISO 2631
compared to the r.m.q. method
Several variations of the The r.m.s. method was found to
Wikström et al. To compare different methods r.m.s. and vibration dose correlate best, but no large
(1991) to evaluate mechanical shocks methods (i.e. exponent differences to r.m.q. or VDV
values from 2 to 10) (exponents 2 and 4) were found
Several methods and The ISO 2631 standard from
variations (in total of 20) from 1978 was concluded to have the
To study discomfort caused by
Fairley (1995) different standard versions of best method: r.m.s. averaging for
tractor vibration
ISO 2631, AFNOR, BSI, and all seat translational axes and
INRS r.s.s. for combining the axes
A review of several studies
Mehta and Several methods used to the No exclusive or single model was
relating to seating discomfort
Tewari (2000) other studies found to be better than others
for tractor operators
1) The r.m.s. or VDV methods
To developed and compare are not recommended for
Alem et al. The r.m.s, VDV and jolt (later
new methodogy for assessing assessing repeated shocks, 2) a
(2004) called ISO 2631-5) methods
repeated shocks new method (ISO 2631-5) was
proposed and found better
1) The vertical seat axis gave
All methods from ISO 2631-
To compare different ride only reliable prediction to
1, BS 6841, AAP and VDI
Els (2005) comfort methods in practice discomfort, 2) no differences
2057 (see later Sections for
using military vehicle were found between the standard
details)
methods
26
2.2.1.4 Issues relating to the understanding of discomfort effects from whole-
body vibration exposure
The perception and comfort studies, which are the basis for our current knowledge,
were made using a single-axis or dual-axis vibrator and using single frequency
(sinusoidal) signals at any time (Hacaambwa and Giacomin 2007, Maeda 2005). The
assumption was that the effects of individual axes can be combined using linear
methods. This assumption has not been conclusively proven, thus more research
needs to be done to confirm the effects in a multi-axis environment (Holmlund and
Lundström 2001). There are also other problems relating to the methods. Maeda
(2005) has made a comprehensive list of necessary research still needed in comfort
research. He points out the limitations and lack of verification of the current ISO 2631-1
standard and its subjective methods and values. Maeda (2005) shows that there is
much more research needed to be done in the subjective scaling area of vibration
exposure. There are other methods for determining subjective comfort, which have not
been studied largely. Also the research has concentrated on the vertical axis in a
seated posture and discrete frequencies. Despite the stated lack of knowledge, the
studies done over twenty years ago are still the most informative and conclusive with
respect to whole-body vibration comfort and perception.
No clear link between discomfort and duration of exposure has been found. Early
studies did not find a consensus of a model for evaluating time dependency for
vibration exposure. There are recent studies on time dependency, which have found it
to be more complex and dependent on frequency and amplitude (Griffin 2007). It is
logical to assume some form of time dependency where discomfort will increase as a
function of time. However, no conclusive results have been produced or a model to
estimate it (Griffin 2007).
The standard method suggests combining axes up to twelve axes, but there are no
validated references for deciding which axes to use (see Section 2.2.1.4). Only two
references for 12-axis measurements were found (Griffin 1990, Maeda 2004). The
references showed weighted r.m.s. values for number of machines for all twelve axes.
However, these two references are not enough to conclude the need for measuring the
additional axes, thus more environments should be evaluated.
It is not yet clear how shocks contribute to the overall opinion, although it can be
hypothesised to have a significant effect. Sudden large shock in the vibration is known
to cause more discomfort, but this varies a lot depending on the control group's gender
27
and age. There have been studies evaluating best methods for shock type vibration
(Alem, et al 2004, Boileau, et al 1989, Els 2005, Griffin 1998, Lewis and Griffin 1998,
Wikstroem, et al 1991), but no clear conclusions of the best overall method is given.
Other characteristics of vibration have been also been investigated, but they have
proven to be more difficult than the amplitude (e.g. gender). Women tend to perceive
the vibration as more unpleasant than men (Griffin 1990), although in earlier studies no
statistical differences were found between the comfort contours (Griffin, et al 1982).
However differences have been found in frequency responses between different body
weights and between men and women (Kumar 1992). This could give a reason to
hypothesise that body characteristics have an influence on vibration discomfort, thus
women and men should have different opinions (Griffin, et al 1982). This has been
concluded by later studies as well (Mansfield, et al 2001). When the vibration comfort is
evaluated several variables can affect the results (example amplitude, duration,
frequency components, sound, temperature, age, gender, personality, etc). This means
that it is difficult to precisely evaluate the discomfort for an individual, but some
statistical estimates can be given if enough information is available.
Rarely is vibration the only source of discomfort. In practice a human will be exposed to
several factors simultaneously and discomfort is a combination of all the factors. There
is little knowledge on multi-factorial effects on comfort (Kolich 2008). Most likely the
dominant factor will be the basis for the opinion, but the relationship of a factor to
discomfort in this type of environment is hard to evaluate. Most studies have been
conducted in a laboratory, where other factors have been minimised. This might give a
wrong indication of the effects of example vibration to discomfort if noise is also
present. There are few studies that have addressed this issue (Fleming and Griffin
1975, Sato, et al 2007).
2.2.2 Health
When the human body is exposed to vibration the body reacts to it negatively (Pope, et
al 2002). The effects can be partly controlled by using the body’s muscles and postural
adaptation, like in skiing or riding a snow mobile. However, the human body is not
evolved for accepting vibration exposure and in the long run the vibration can cause
health problems (Bovenzi, et al 2002). If posture is not optimal (e.g. back twisted) the
harmful effects of vibration are amplified (Wilder and Pope 1996). Other factors that
might advance the harmful effect are lifestyle habits (i.e. smoking, obesity), body’s
measurements (height, weight), duration of exposure (Mansfield 2005, Teschke, et al
28
1999), although there are inconsistent results of linking these factors to back pain
(Hostens, et al 2004, Hoy, et al 2005). The current understanding is that a moderate
vibration exposure alone is not unhealthy, but it becomes harmful if added with other
negative factors or if there is already a problem with the back.
2.2.2.1 Low back pain
It has been long known that vibration exposure has a link to health problems, most
particularly to low back pain (LBP) (Palmer, et al 2000). This has been concluded
based on epidemiological studies (Bovenzi and Hulshof 1998, Seidel and Heide 1986),
field studies (Mansfield 2003, Paddan and Griffin 2002b), medical reports (Pope, et al
2002) and ergonomic studies (Cole and Grimshaw 2003, Hoy, et al 2005). LBP is one
of the most common work related health problems in the world and causes billions of
Euros worth of costs to countries and companies (Deprez, et al 2005, Hostens 2004).
There is a consensus between researchers that many back pain cases are related to
vibration exposure either directly or indirectly (European Comission 2007).
LBP has been studied a lot in the recent years (Bovenzi and Betta 1994, Bovenzi, et al
2002, Bovenzi, et al 2006, Hostens, et al 2004, Lings and Leboeuf-Yde 1998). Many
epidemiological studies have concluded a link between the vibration characteristics and
exposure period and subject characteristics to the prelevance of back pain. Studies
have been normally cross sectional or questionnaires either in a smaller community
(e.g. harbour workers; agricultural workers) (Bovenzi, et al 2002) or country wide
(Rehn, et al 2005a). Literature reviews have gathered results and conclusions of these
studies to find any similarities (Lings and Leboeuf-Yde 1998, Seidel and Heide 1986,
Shoenberger RW 1984, Stayner 2001, Teschke, et al 1999).
Based on the studies it has been suggested that some LBP cases are caused by
vibration exposure (Bovenzi and Hulshof 1998, Seidel and Heide 1986). Numerous
back disorders can be involved, such as sciatica, lumbago, general back pain,
intervertebal disc herniation and degeneration (Pelmear and Leong 2002). Seidel and
Heide (1986) concluded based on the large health data that whole-body vibration has
an effect increasing the health risk. If heavy lifting is additionally involved in the work
period, then there is even greater risk of back pain injury. Dose period seems to have
an effect as well as certain frequencies and amplitudes. Although LBP and vibration
has not been clinically linked together it is a statistical fact that LBP is associated with
high level vibration exposure.
29
Bovenzi et al (2006) conducted a large study with almost 600 Italian workers to
investigate prelevance of LBP in professions with whole-body vibration exposure. The
study recorded detailed personal and work related data from the workers using
questionnaires and measured several machines and work environments. The study
found that occupations with whole-body vibration exposure were significantly
associated with occurance of LBP. It also found that number of work years correlate
with prelevance of LBP.
2.2.2.2 Other effects
Back pain is not the only vibration-related health problem. Research has been
completed on other health effects of vibration (Griffin 1990, Ishitake, et al 2002, Rehn,
et al 2004, Seidel and Heide 1986, Shoenberger RW 1979, Sommerich, et al 1993).
The effects can be for example digestive disorders, hearing damage, neck pain
(Magnusson and Pope 1998). The neck and shoulders have been considered in some
studies (Rehn, et al 2002, Rehn, et al 2005a), although no definite relation has been
proven or systematic conclusions produced. Vibration has also an effect on internal
organs, for an example to gastric motility. A study (Ishitake, et al 2002) showed that
vibration exposure, especially at low frequencies, has, even at short term, an effect to
gastric motility. Even though the effect might not be strong for healthy persons, it
shows that internal organs are sensitive to vibration, especially at their resonance
frequency. Other problems might be sciatica, genitourinary problems and hearing
damage (Griffin 1990). Vibration might cause changes also to heart-rate and
respiration, however there are not enough studies for any firm conclusions. Based on
the findings it can be speculated that a part of the comfort feeling can be produced
from the body’s reaction. This hypothesis has not yet been proven.
2.2.2.3 Issues relating to the understanding of health effects from whole-body

vibration exposure
Epidemiological studies are based on questionnaires and statistics. In vibration

exposure the goal has been to evaluate the statistical factor of vibration causing LBP.
The publications have made cross-sectional studies with other factors and tried to
determine the percentage that vibration exposure might have caused (Rehn, et al
2002). Only few studies have focussed on the effect of dose period to the prelevance of
back pain (Hostens, et al 2004). Because symptoms might not occur immediately, the
vibration will cause problems in the long run, which in worst case scenario might be
irreversible (Hulshof, et al 2002). It is not clear how exposure duration is linked to the
30
health problems, but it is logical that longer durations increase the likeliness of the
problems (Bovenzi, et al 2006). This is also the basis for the standardised methods,
which have a time factor linked to the exposure value (ISO 2631-1 1997).
The epidemiological studies have had tens of thousands of people participating in the
studies, but in critical reviews of the publications lots of problems have been found with
the data and study procedure, thus not all publications have been found as good
quality (Seidel and Heide 1986). The large number of studies has been unable to show
that the vibration exposure and LBP is conclusively connected. Also there is not
enough evidence of effect of the vibration exposure duration and LBP (Seidel and
Heide 1986, Teschke, et al 1999). It is very hard to separate LBP caused by the
vibration and other factors such as, lifting, ergonomics, lifestyle habits, weight, body
structure (Griffin 1990, Teschke, et al 1999).
2.2.3 Motion sickness
Although motion sickness is related to discomfort and perception, the cause of it is

outside the scope of this thesis. Motion sickness is believed to be produced by the
conflict of body’s different sensory systems and has specific symptoms, like nausea
(Mansfield 2005). The motion sickness is separated from health and comfort effects by
its limited frequency range (0 to 0.5 Hz) (ISO 2631-1 1997). There are many variations
of motion sickness, such as sea sickness, car sickness, etc, but the basis for all these
variations are believed to be the same.
2.2.4 Summary
It has been shown that frequency, amplitude, location and duration of exposure have
an effect on the comfort and perception of vibration. The equivalent comfort contours
have been shown by many studies to have similar pattern, where lower frequencies
should be emphasised. Discomfort is formed based on a combination of axes where
vibration is present. It has been shown that axes should be combined for better
correlation (Parsons and Griffin 1983). However, it was found that little research has
been made since the introduction of the standard method to validate and improve the
found results.
Similar to discomfort, several results for health effects have been concluded. It is
strongly believed that vibration exposure can cause health problems in the long run
(most commonly low back pain). Many studies have shown results implying this, but
still there is no proven model to predict and assess health risks from the whole-body
31
vibration. The position of the lumbar spine is very important for preventing back
problems, thus ergonomics greatly influence the prevalence of the back pain (Hoy, et al
2005). This and other factors, such as lifting, make it extremely difficult to separate the
effects from vibration exposure in health problems.
Even though a large number of studies have been conducted in this research area
since the 1930’s, there are still many issues that are not yet understood. Only a
relatively small number of studies have contributed to the understanding of the effects.
There have been few studies which have tested the ISO 2631-1 (1997) method in a
multi-axis environment and even fewer which have used field-like stimuli. Additionally it
has been rare that other than seat translational axes have been measured from work
machines, even though it is not clear how the additional axes will contribute to a
human’s perception.
Whole-body vibration research has focussed generally on evaluation of health effects.

Although health is usually the primary concern, the knowledge behind it is based
mainly on comfort and perception studies. The hypothesis has been that human
perception is linked to the body’s “health” response. This has been also reinforced by
human biomechanical studies, which have shown resonance frequencies in the same
areas as human perception are the strongest and most clearly in the vertical direction
(Mansfield and Maeda 2005b, Nawayseh and Griffin 2003). The knowledge behind the
current health evaluation method (ISO 2631-1 1997) is based mainly on studies of
human perception and comfort (Griffin, et al 1982, Parsons and Griffin 1978a, Parsons,
et al 1978, Parsons and Griffin 1980, Parsons and Griffin 1982, Parsons, et al 1982).
Earlier studies tested the perception thresholds of several persons to different
frequencies and amplitudes. The results from those studies have formed the basis of
the methods in use today for assessing health and comfort effects.
2.3 Methods for understanding the effects from vibration exposure

Before a method to evaluate health or discomfort can be developed, studies need to be
conducted to find characteristics of vibration that cause the effects. This can be done
by either asking subjects how they perceive vibration (i.e. subjective evaluation) or by
measuring human body’s dynamic response (i.e. biomechanical evaluation).
Methods for finding how humans react to vibration over a specific frequency range
have been conducted since the 1930’s (Oborne 1977). Most of the earlier studies used
a sinusoidal stimulus (e.g. 3 Hz) and asked subjects to adjust the stimulus until they
considered it to be “comfortable” or “perceptible”. More recent studies have also used
32
different methods, but are still based on laboratory environments and sinusoidal or flat
spectrum stimuli (Maeda 2005).
Because of limited possibilities to directly measure physical effects from vibration on

the human body, subjective methods have been mainly used for understanding both
health and discomfort (Mansfield 2005). Perception and sensitivity to certain vibration
characteristics indicate how humans sense different frequencies of vibration. Even
though biomechanical studies can give a good indication of dynamic responses of body
parts, which can be linked to most harmful vibrations, it does not improve the
understanding of the psychological effects of vibration (e.g. threshold studies can find
out the lowest levels of vibration at each frequency, which human can detect). It is also
practically impossible (and unethical) to study how long exposure or high amplitude will
cause permanent damage to the human body. The more ethical approach is to use the
subjective methods.
There are various experimental methods used for comfort and perception research
(Figure 11). The purpose of using the methods is to improve understanding on how the
vibration characteristics link to the subjective feeling. This includes also developing a
scale that can be used to evaluate comfort feeling based on the vibration
measurements. One example of a comfort scale is presented in Annex C of the
standard (ISO 2631-1 1997) (see Table 6).
The psychophysical methods used for vibration related issues can be divided into two
categories: 1) constant measurement and 2) subjective scaling method (Maeda 2005).
The constant measurement method has been used mainly for concluding the threshold
of a human sense to characteristics of the vibration. The method uses a pre-
determined scale for linking the subjective results. The subjective scaling method on
the other hand has been used to find out the subjective scaling compared to the
measured vibration value and it includes creation of a scale (Maeda 2005). Because
the methods have a different approach they also will have different results. It is thus
important to select the right method for the purpose.
33
Psychometrics
Constant Measurement Methods Subjective Scaling Methods
Constant Method Interval Proportional

method of Limits Scale Scale
Adaptive
Method of Category Paired Magnitude
Psychological
Adjustment Judgment Comparison Estimation
Method
Figure 11. Examples of psychophysical methods for studying comfort and perception from
vibration exposure (adapted from Maeda (2005)).
2.3.1 Constant measurement methods
It is advantageous to know the lowest levels of vibration that can be perceived at

different frequencies, postures, locations and axes. Many studies have used this type
of method in research (Fothergill 1972, Fothergill and Griffin 1977, Jones and
Saunders 1974, Maeda 2005, Oborne and Clarke 1974).
2.3.1.1 Constant method
Where stimulus pairs presented by preliminary experiment are presented several times
in random order (Havelock, et al 2008). Subjects are asked to judge each pair. The
method takes a relatively long time as many pairs need to be judged.
2.3.1.2 Method of limits
A stimulus is presented by varying the stimulus parameters (i.e. frequency or

amplitude) in a fixed step from low to high (Havelock, et al 2008). Subjects will
compare stimulus pairs and judge which stimulus is lower and higher. The procedure is
predetermined by the measurer.
2.3.1.3 Method of adjustment
Method of adjustment has been popular for finding out the vibration thresholds (Maeda
2005). The method can be described as: “where subject adjusts the level of a standard
34
frequency of vibration to have an equal subjective magnitude of another vibration” –
Kuwano, S. (Handbook of Signal Processing in Acoustics). This technique can be used
to produce comfort contours in term of equivalent comfort.
2.3.1.4 Adaptive psychological method
Differing from aforementioned methods, in this method stimulus is changed depending

on how subjects react to it (Havelock, et al 2008). It will require a computer to conduct
the measurements and activity from the measurer. There are several methods of
adaptive procedure (e.g. “up and down method”).
2.3.2 Subjective scaling methods
Subjective scaling methods can be further divided to interval and proportional scale
methods. The aim is to find subjective scaling to subjectively perceived quantity
(Maeda 2005).
2.3.2.1 Magnitude estimation method
The Magnitude Estimation (ME) method is a scaling method for obtaining proportional
scale of vibration. The method has been proposed by Stevens (1957). It can be
suitable for determining if evaluation index corresponds to the subjectively perceived
quantity. The goal is to link psychophysical magnitude of stimulus to a physical
magnitude. Exponent n is found using test subjects and a seat cushion with specific
stiffness. The method can be formulated as:
ψ = kφ n (1)
Where ψ is psychophysical magnitude of a stimulus, φ is physical magnitude of a

stimulus and k is constant and n is the growth of sensation exponent.
The method includes assessing a stimulus to a reference stimulus and judging the
relative effect by giving a number relative to the reference stimulus. Normally a fixed
reference stimulus is used (e.g. 5 Hz and 1 m/s2 r.m.s.) and it is assigned as “100”.
Figure 12 shown previously by Morioka and Griffin (2006) shows results using the ME
method.
Improvements to ME method has been proposed by Howarth and Griffin (1988). The
model takes into account both static and dynamic properties:
ψ = a + bφ s n + cφv n
s v
(2)
35
Where ψ is psychophysical magnitude of a stimulus, a, b and c are constants, φ s and
φv are seat stiffness and vibration magnitude, and ns and nv are exponents determined
by the rate of increase in discomfort associated with the stiffness and vibration
magnitude.
Ebe and Griffin have used it for determining overall seat discomfort (Ebe and Griffin
2000b). At low vibration magnitudes the static properties of the seat dominate the
comfort, but when magnitudes are increased the effect of dynamic properties have
more influence (Figure 12).
Figure 12. Ebe’s model for predicting seat discomfort using static and dynamic properties (Ebe
and Griffin 2000a).
Ebe’s model did not include a time factor, which can be assumed to have an effect.
Later Mansfield (2005) proposed an improved model, where also time was included
(Figure 13). The equation is improved by including temporal factors:
ψ = s s + f t t + d v a + itv ta (3)
Where ψ is the rating of discomfort, ss is the static discomfort constant, ft is a fatigue

constant, dv is the vibration discomfort constant, itv is an interaction variable, t is the
time (mins) and a is the r.s.s. acceleration (Mansfield, et al 2007).
36
Figure 13. Improved model for predicting seat discomfort using static, dynamic and temporal
factors (Mansfield 2005).
2.3.2.2 Difference threshold
The difference threshold is a method for obtaining the amplitude of two vibrations that
can be just perceived as different. The importance in knowing the human perception for
changes is that the improvement done for an example to car suspension is enough to
be noticed as improved ride comfort. Usually, difference thresholds are determined
using the UD (The Up-and-Down) method or the UDTR (The Up-and-Down
Transformed Response) method.
2.3.2.3 Paired comparison method
In this method chosen stimuli are presented to the subjects as pairs and the subject will
select a better one. Each combination of stimuli is paired and a matrix of proportion of
time when one stimulus is preferred over another is created. The ranking of stimuli is
calculated based on the matrix (Oborne 1978b).
2.3.2.4 Category judgment method
The category judgment method (Guilford 1954) seems to be good for linking a
subjective phrase or word to the vibration characteristics (Maeda, et al 2008). The
method implies that a reaction to a stimulus forms a normal distribution on a
psychological continuity. The procedure for creating a scale starts with test subjects
evaluating different stimuli using a pre-determined scale of phrases from “not
uncomfortable” to “very uncomfortable”. The results are then summarised in a table by
the categories representing the phrases and different stimuli. Each phrase represents a
number scale from 1 to 5 so each stimulus will have a distribution of subjective
responses. The table is converted to a relative frequency table, which is then converted
37
to a normal distribution for each stimulus. The normal distribution results are then used
to produce an average scale for each stimulus. Kaneko et al. (2005) have explained
this procedure in their study.
2.3.2.5 Borg scale
The Borg scale has been used in some occasions as it combines ease-of-use of a
category scale with the analytical flexibility inherent in numbers reported using a ratio
scale. It has been widely used for fields of psychology, physiology and ergonomics to
rate sensation of pain, fatigue and discomfort (Hacaambwa and Giacomin 2007).
2.3.2.6 Continuous judgment method
The continuous judgment method allows test subjects to evaluate feelings in real-time
and the judgments can be correlated to the stimuli by averaging values in different time
windows. The method has been developed for evaluating continuous impressions of
sounds which vary over time (Kuwano and Namba 1985, Namba and Kuwano 1988,
Namba and Kuwano 1979). The are several variations of the continuous judgment
method: 1) by category, 2) by line length and 3) by selected descriptions.
Continuous judgment by category involves a predetermined category (e.g. “loud”, “very

loud”). Subject will judge sounds by notifying the feeling continuously relating to any
category. There is no need to notify feeling unless it changes. Normally a keyboard or
buttons with the categories are used to allow easy way of judging.
Continuous judgment by line length allows subjects to express more subtle changes
that could be too small to represent using predetermined categories. Thus cross-modal
matching of sounds to line length can be analysed. There needs to be no fixed scaling
or numeral for the line length and only the extremes are noted to the subjects (e.g. “no
discomfort” and “high discomfort”).
Continuous judgment by selected description can be used to evaluate multi-

dimensional impressions. This includes a list of adjectives which describes the stimuli.
Example in sound studies the adjectives could be “beautiful”, “loud”, “soothing” and so
on.
The continuous judgment method is developed for judging long term stimuli using
instantaneous judgments in situations similar to everyday life. One of the merits of the
method is that results can be easily stored in a computer for analysis in various
viewpoints later. There is an inherent lag between reaction and stimulus. There has
38
been studies investigating the reaction time for sounds (Kuwano and Namba 1985). It
was concluded that an average 2.5 second delay was found. Even though the method
has not been extensively used for evaluating discomfort from vibration, it can be
assumed to work similarly as in the sound research. This is because the method was
developed to understand subjective judgments of humans, not particularly noise
characteristics, thus can be used basically with any factor.
2.3.3 Discussion of the methods
Each method can provide results and conclusions in its own limited scope. If only one
type of method is used, then the results will be impossible to validate outside the scope
of the research. For example the current subjective assessment table in the standard is
based only on one method, thus it is difficult to evaluate the scope of the table for
predicting discomfort in practice. Each method also has specific assumptions and
procedures to be used correctly, thus not all methods can be used for example in field.
Some methods have been criticised as biased (e.g. method of limits), while others have
not been rarely used, thus does not have high confidence levels (i.e. vibration
greatness) (Havelock, et al 2008).
The ME method has been most popular for finding comfort contours and correlation to
discomfort. However, it has been also criticised (Hacaambwa and Giacomin 2007). The
method requires a reference stimulus for comparison and the selection of the reference
stimulus will affect the results. It is most practical to use reference stimulus before each
studied stimulus, so that subjects will have short time duration between the stimuli thus
still remembering the reference. Normally the stimuli have been short (i.e. few seconds)
and sinusoidal or random flat band. The method also assumes that if a subject doubles
the judgment number relative to the reference stimulus, then discomfort is also
doubled. There are studies using the method in the field (Fairley 1995). In the field, the
first object has normally been the reference stimulus and the rest of the objects are the
studied stimuli. In this case the subject might have hard time of remembering anymore
the reference stimulus. In an environment where there is no fixed reference stimulus
available (i.e. in real life) the method is hard to use.
Only a few studies have used the category judgment method. Maeda et al. (1983) used
it for producing a scale to evaluate localised vibration transmitted to hand and arm and
Sumitomo et al. (1998) for identifying changes of subjectively perceived values of
vibration of passengers of Shinkansen bullet trains before and after Kobe Earthquake
of 1995. Kaneko et al. (2005) did a study of linking different vibration exposures of
39
different frequency content to a subjective scale (Figure 14). The results showed that
the scale was different for the same r.m.s. vibration level when the frequency content
was different. The figure also shows the overlapping of ISO 2631-1 scale. The opinion
of the authors was that using the category judgment method the subjective evaluation
of vibration and physical value could be clarified. This can be true, but the method and
the results show that different scales should be created for different vibration
exposures.
Figure 14. The difference of subjective scales compared to ISO 2631-1 scale and using
category judgment method (Kaneko, et al 2005).
The continuous judgment method has been used in many studies regarding noise
annoyance (Kuwano and Namba 1980, Kuwano and Namba 1985, Namba and
Kuwano 1980, Namba and Kuwano 1982, Namba and Kuwano 1988, Namba and
Kuwano 1979). However there are no references for using the method relating to
vibration. Based on the information on the method, it seems suitable for field and
evaluating discomfort of long exposure. It can be used to evaluate instantaneous
discomfort for an undetermined stimulus. Because of lack of studies in vibration, it
needs a validation study before it can be used.
40
Other methods have been also used, which were not discussed in this review. For
example Oborne (1977) used a paper with a line drawn and subjects evaluated
discomfort by marking a cross in the line. No specific category or scaling was given,
thus the distance of the mark from the beginning of the line was used to form a number
corresponding to discomfort. This type of approach has been popular in field studies.
In addition to selecting methods it is important to understand the role of the subjects.

Already in 1976 a criticism of the study approaches for finding the effect of vibration on
subjects was discussed (Oborne 1976). Many studies have been conducted only in
laboratory and using students or staff, not “fare paying” customers. Test subjects will
expect certain vibration levels and know that they are in a test. Their expectations are
different from real customers.
The usability of the methods to evaluate vibration related discomfort is more accurate
in an environment where vibration is the dominant factor or where other factors have
been minimised. In these environments it is possible to assume that vibration is the
primary cause of discomfort. In other environments the usability of the methods are
decreased as they tend not to include any other factors (e.g. noise) in the analysis.
This is in effect a limitation of field trials.
2.3.4 Summary
Several methods exist for evaluating comfort and perception from vibration. The
methods have been derived from psychology studies, thus will require some knowledge
in that area. It is important to select a correct method when studying effects from
vibration, as each method has its own merits, limitations and scope. It is thus equally
important to validate the findings using different methods to improve the validity of the
results. Several methods were evaluated, which gave an understanding of possibilities
for evaluating discomfort in field environments. It can be concluded that only few
methods can be considered for the approach chosen in this thesis.
2.4 Methods for evaluating the effects of vibration exposure

The principal measurement and analysis procedure for evaluating health and comfort
from vibration exposure is described in the standard ISO 2631-1 (1997). It is based on
several studies conducted on subjects in a laboratory. Also other methods exist, which
can be used to evaluate effects, mainly relating to health (i.e. physical effects). Some
methods can be used for discomfort evaluation as well, but beyond frequency weighted
41
r.m.s. method in ISO 2631-1 there are no validations of their applicability to discomfort
(Els 2005).
Four main methods exist for objectively evaluating ride comfort (Els 2005): 1) ISO
2631-1 (1997), 2) BS 6841 (1987), 3) VDI 2057 and 4) Average Absorbed Power
(AAP), but also additional methods exist, which have been used for specific purposes
regarding comfort and health (e.g. Jerk or Spinal response method in ISO 2631-5).
Even though the methods used in this thesis are limited to ISO 2631-1 and BS 6841
standards, it is necessary to understand other methods and reasons for their existence.
This gives insight to different aspects of the vibration effects.
Most methods are based on manipulating acceleration signal measured near or

between the human body and a vibration platform. The purpose of signal manipulation
is to emphasise frequencies and axes that are most harmful to a human body.
Typically a single value is calculated, which can be compared to other environments
and criteria determining the severity. There are some methods which are based on
force input or calculating transient shocks from vibration.
2.4.1 ISO 2631-1 and BS 6841 methods
Two main methods exist for evaluating the effects from vibration: 1) frequency
weighted r.m.s. and 2) frequency weighted vibration dose value (VDV). These are the
most widely used methods for producing results from vibration measurements. Both
methods use same filters (i.e. weighting curves) to process the acceleration data and
multiplying factors for emphasising each axis. The difference is that the r.m.s. method
has second power averaging while VDV uses fourth power without averaging duration.
Thus VDV emphasises more shocks than the r.m.s. method. Additionally BS 6841
introduced a r.m.q. method, which is similar to the r.m.s. method, but using fourth
power averaging. It is also comparable to the VDV value, if exposure duration is
known, thus linking both the main methods.
The methods described here are from ISO 2631-1 (1997) and BS 6841 (1987)
standards, which are discussed in more detailed in Section 2.6.
2.4.1.1 Root mean square (r.m.s.)
R.m.s. method is a statistical measure of the magnitude of a varying quantity. It is

especially useful for calculating mean of values which are both positive and negative
(e.g. acceleration signal), as the method weights each value as positive. R.m.s. is
calculated as:
42
1
1T 2 2
a w (T ) =  ∫ a w (t )dt  (4)
T o 
Where aw(t) is frequency weighted acceleration (m/s2) at time t and T is time (s).
2.4.1.2 Root mean quad (r.m.q.)
R.m.q. is similar to the r.m.s. method, except each value is calculated to the fourth
power, thus higher values are emphasised more (e.g. shocks in vibration):
1
1 T
 4
a w (T ) =  ∫ a w (t )dt 
4
(5)
T o 
Where aw(t) is frequency weighted acceleration (m/s2) at time t and T is time (s).
2.4.1.3 Vibration dose value (VDV)
VDV is similar to r.m.q., but it includes a time dependency. VDV value is calculated
from the frequency weighted acceleration data using the following equation:
1
T 4
VDV =  ∫ [a w (t )] dt 
4
(6)
o 
Where aw is frequency weighted r.m.q. value at time t and T is time of measurement

period (s).
2.4.1.4 SEAT value
The SEAT value can be used to determine how well does a seat attenuate vibration
from the floor. The SEAT value (%) is calculated by comparing frequency weighted
r.m.s. (or VDV) values from the seat surface and floor and is supported by the standard
method (ISO 10326):
awseat
SEAT = ⋅ 100 (7)
awfloor
Where awseat and awfloor are frequency weighted r.m.s. values from floor and seat for
each axis.
Even though some studies have used the SEAT value as a predictor for dynamic seat
comfort (Van der Westhuizen and Van Niekerk 2006), it is designed and mainly used
for evaluating characteristics of a seat damping. Paddan and Griffin (2002a) conducted
43
a large study (100 vehicles) to evaluate the effectiveness of seating to minimise whole-
body vibration. The results showed that in most machines the seating attenuated
vibration, but could be even more improved with another seat. It is likely that a smaller
SEAT value (i.e. better attenuating) can suggest also improved comfort, but it most
likely cannot be used as a single predictor.
2.4.2 Other methods
2.4.2.1 Vibration greatness
Vibration greatness (VG) is a method derived from acoustics (Miwa and Yonekawa
1974). The basis of the method is to use either octave (i.e. simple method) or one-third
octave (i.e. complex method) bands to produce a single value that corresponds to the
loudness or to the loudness level of a given sound. Vibration greatness uses the same
principle for evaluating the effect of vibration to humans as for loudness. Vibration
greatness level (VGL) corresponds to the loudness level and vibration greatness (VG)
corresponds to the loudness. First a subjective threshold values for the vibration has to
be determined. This can be done, for example, using a paired comparison method
where a test stimulus is compared to a fixed stimulus (e.g. sinusoidal vibration). Based
on this procedure vibration acceleration level (VAL) of each band can be obtained. The
VAL values can be turned into VG values using a pre-determined table. Example of a
table can be seen at Miwa’s publication (Miwa 1969). Now the VGT value can be
calculated using equations 8 or 9 depending on which method is used:
VGT = VG M ,1 + 0.3((∑ VGi ,1 ) − VG M ,1 ) (8)

i
VGT = VG M ,1 / 3 + 0.13((∑ VGi ,1 / 3 ) − VG M ,1 / 3 ) (9)

i
Where VGM,1 is largest VG value of all VGi and VGi is a VG value of a frequency band.
Suffix 1 and 1/3 denotes the method used (octave or one-third octave). The VGL value
can be obtained from VGT again using a table. The VGL value can be used to estimate
subjective response of a specific vibration stimulus.
2.4.2.2 Jerk
Jerk (symbol j) is a term for third derivate of position, which also has an influence on
comfort (Speckhart and Harrison 1968). Jerk has been studied and used for vibration
control and ride comfort evaluation of trains, elevators, robots, and so on. Jerk is
44
important when evaluating the destructive effect of motion on a mechanism or the
discomfort caused to passengers in a vehicle. The movement of sensitive equipment
needs to be kept within specified limits of jerk as well as acceleration to avoid damage.
When designing a train the engineers will typically be required to limit the level of Jerk,
because of passenger comfort.
2.4.2.3 Spinal response method of ISO 2631-5 for evaluating effects on

multiple shocks
A random vibration exposure, which is experienced in real work environments, includes

transient shocks as well as other kinds of vibrations. Even though aforementioned
standardised methods include all vibration characteristics in the analysis, they tend to
undervalue the effects of multiple shocks as they average the vibration data. For this
reason additional methods have been developed, which separately evaluate the effects
of shocks.
An addition to the ISO 2631 standard series was produced, which analyses the effects
of shocks to health (ISO 2631-5 2004). Although the method is not directed at comfort
evaluation (only health), it shows that vibration effects can be assessed also using
methods other than frequency weighting and time dependency. The ISO 2631-5
method could be a basis for an additional method to assess shocks and their relation to
comfort.
The purpose of ISO 2631-5 is to define a method for analysing the effect of multiple
shocks in relation to human health. The standard is based on a lumbar spine response,
because it is believed to be affected by shocks the most. The standard addresses the
human exposure to multiple shocks only when the human is seated.
To evaluate health effects the standard introduces static compressive stress value
(Sed). Sed is calculated from the sixth power sum of acceleration dose values
multiplied with dose coefficients. If Sed value is below 0.5 Mpa there is a low probability
of adverse health effects at lifetime exposure. Above 0.8 Mpa there is high probability.
Results between 0.5 and 0.8 Mpa indicate moderate health effects at lifetime exposure.
There are separate procedures for horizontal and vertical directions in the standard.
Horizontal directions are assumed to have a linear response.
2.4.2.4 VDI 2057
VDI 2057 is a German standard, which is used for quantifying ride comfort. The first
version was published in 1963 (Els 2005). Since then it has become more similar to
45
that of ISO 2631-1 series (1985 version), but has a K-factor for comparison of
subjective sensations of humans. The acceleration time signal is converted to
frequency domain and r.m.s. values are calculated for each 1/3 octave center
frequencies. K-values are used for weighting each frequency:
1 ≤ f ≤ 4 Hz K z = 10 ⋅ a z ⋅ f , 4 ≤ f ≤ 8 Hz K z = 20 ⋅ a z , 8 ≤ f ≤ 80 Hz K z = 160 ⋅ a z ⋅ f
Where f is the frequency, az is amplitude in m/s2 for vertical direction and Kz is a

weighting factor for each frequency. The values can be calculated only for vertical
direction and for a seated person. The resultant plot is compared with limit curves like
in ISO 2631-1 (1985).
2.4.2.5 Averaged Absorbed Power (AAP)
The AAP method was developed in 1960’s by the US military (Els 2005). The absorbed
power can be calculated as:
N
AP = ∑ K i ⋅ ( Airms ) 2 (10)
i =0
Where AP is absorbed power, Ki is a weighting factor for i 1/3 octave frequency, N is

number of frequencies, Airms is the averaged power for 1/3 octave frequency.
2.4.3 Discussion of the methods
Even though several methods exist for evaluating health and discomfort, in practice the
r.m.s. and VDV methods have been mostly used, and there is only adequate
information on assessing discomfort using the r.m.s. method. The VDV method might
be better if comparison of exposures of different durations are to be compared or
shocks are involved. The r.m.q. method has similar effect, thus can be regarded as
better option if direct comparison to the r.m.s. values is required.
In a study it was concluded that a SEAT value could be used as a reliable metric to
select the best seat for each vibration characteristic (Van der Westhuizen and Van
Niekerk 2006).
Also the VG method has been rarely used in whole-body vibration research. Miwa
(Miwa 1969, Miwa 1967) used it to determine perceptional thresholds and equal
sensation characteristics for random vibrations for vertical and horizontal directions.
More recently Maeda has used the VG method and also has evaluated the usability of
46
the method compared to other methods (Maeda 2005). He concluded that VG method
works well for single-axis stationary vibration, but not for multi-axis environment.
Jerk is an important factor when elevator speed control or train vibration control is
designed, and it is reasonable to assume that it might relate to whole-body vibration
comfort. There are no systematic studies concerning jerk and whole-body vibration
comfort and it has not been standardised.
ISO 2631-5 was produced based on the research carried out for the U.S. military
(Alem, et al 2004). The basis of the research was that in 1980’s there were already
reports indicating health effects of vibration exposure in military vehicles, even though
these vehicles passed existing whole-body vibration standards. There is no indication
of whether ISO 2631-5 methods correlate with discomfort judgment.
Els (2005) has made a test for comparing four comfort evaluation methods (BS 6841,
ISO 2631-1, VDI 2057 and AAP) using military vehicles. He also used unweighted
r.m.s. values as comparison. The measurements were not conducted directly by the
procedure of the standard (no backrest or feet measurements), but rotational directions
were calculated. The results showed that all four methods exhibited some consistency
in evaluating the comfort to vertical vibration, but not to other directions. The study had
limited scope and implied that the other methods (VDI 2057 and AAP) did not provide
any new insight, thus there is no motivation to use them.
2.4.4 Summary
There are methods for evaluating discomfort from vibration, but most of them have not
been tested or validated in practice. This is also a problem with the r.m.s. method
proposed by ISO 2631-1 standard, because only few studies since 1997 have tested
the method (Maeda 2005). The biggest problem is that the knowledge in the standard
is based on single-axis studies and it is a good probability that human behaves
differently in a multi-axis environment (Holmlund and Lundström 2001).
2.5 Measuring vibration

Evaluation of the effects from whole-body vibration requires measurement of the
vibration in the form of acceleration. Based on the standards the vibration is measured
between a human body and a vibrating platform. In whole-body vibration this generally
means a seated person with sensors between seat backrest, cushion and floor. There
are guidelines for measurement periods as well as installation of the sensors. Different
number of sensors might be required depending if health or discomfort is evaluated.
47
2.5.1 Requirements
There are three standards that define the technical requirements, analysis methods
and installation procedure for whole-body vibration. ISO 2631-1 (1997) is the main
standard for whole-body vibration. It defines the method and basic guidance for
conducting the measurements and analysis. ISO 8041 (2005) defines the guidance to
develop and test the equipment for conducting the measurements and ISO 10326
(1992) defines the installation of the seat sensor more specifically (ISO 10326 1992).
However none of the standards define technical or installation specifications for
realising 6-axis seat sensor, which is required for the full evaluation procedure.
The technical requirements for measuring the 12-axis vibration are: 1) a 6-axis seat
sensor, 2) one 3-axis sensor each for backrest and floor, 3) a data acquisition system
to record the sensor signals and 4) software to calculate and analyse the results. The
6-axis seat sensor is needed to record three translational (fore-and-aft, lateral and
vertical) and three rotational (roll, pitch and yaw) directions. Three translational
directions are recorded from the floor and backrest (Figure 2; 12-axis coordinate
system). All of the twelve axes need to be analysed as acceleration (m/s2 for
translational and rad/s2 for rotational directions). This means that either acceleration
sensors are used or the recorded data needs to be converted to acceleration data (e.g.
if inclinometers or force sensors are used).
There are no specific requirements for the data acquisition process, except that the
accuracy and tolerance of the whole equipment must be within the requirements of the
ISO 8041 standard. The sampling process and filtering for the frequency weighting can
be realised in various ways. In practice 256 Hz sampling rate and sensor sensitivity of
0.01 m/s2 are the minimum requirements.
Because the standards do not specify either mechanical or electrical characteristics or

the detection method for the 6-axis sensor, it can be designed in various ways.
However the selection of the sensor type and detection method is crucial for
developing the whole equipment, because it also defines the other components and
their requirements. Literature search revealed at least three different approaches to
record all six directions:
• 3-axis acceleration sensor for translational directions (m/s2) and 3 inclinometers

for each rotational directions (inclination [degrees]);
• 3-axis acceleration sensor for translational directions (m/s2) and 3 gyros for
rotational directions (angular velocity rad/s);
48
• 3-axis acceleration sensor for translational directions (m/s2) and 3 acceleration
sensors for detecting rotational directions (rad/s2).
One of the practical limitations for realising the 6-axis sensor is the physical size. The
sensor needs to be thin enough to be sat on and small enough to fit under a person’s
buttocks. This limits the height and width to about 150 x 200 mm, and thickness to 20-
25 mm. Another limitation is the frequency range. To measure whole-body vibration the
sensors should allow analysis of frequencies between 0.5 to 80 Hz of acceleration. So
if other than acceleration sensors are used the data needs to be converted to
acceleration for example using derivation, which amplifies noise. For example
inclinometers normally detect frequencies between 0 to 20 Hz, so they are a poor
choice for whole-body vibration measurements. Gyros on the other hand detect higher
frequencies, but are more expensive and complex components than accelerometers.
No single 6-axis accelerometer component was yet commercially available in 2005.
Recently (2009) integrated 6-axis sensors have been introduced, which are based on
measuring translational axes using accelerometer and rotational axes using gyros.
Their suitability for whole-body vibration is not any better than using a number of
accelerometers.
2.5.2 Commercial equipment
Because there is no proper guidance in the standards, only a few have designed and
tested a 12-axis measurement system and developed a 6-axis sensor pad (Whitham
and Griffin 1977, Yamashita and Maeda 2003). An early seat pad design, called SIT-
BAR, made twenty years before the current standard was tested by Whitham and
Griffin (1977). The goal of the study was to assess a method measuring whole-body
vibration on soft seats. In more recent publications, Yamashita and Maeda (2003) have
developed equipment for 12-axis comfort measurements and analysis. Based on this
development commercial equipment has been produced by IMV Co (Figure 15). Only
ISVR at the University of Southampton and IMV Corporation has developed the
equipment for 12-axis comfort measurements and analysis. Neither ISVR nor IMV Co.
actively markets the equipment and in any case the cost would be tens of thousands of
Euros. A more recent paper about six axis measurements of skidders reported a
development of 6-DOF seat pad sensor (Cation, et al 2008), which was a prototype for
research purpose.
49
Figure 15. Commercial 12-axis measurement equipment for discomfort evaluation by IMV Co
(cropped from a product flyer).
It seems that the equipment sold under "whole-body vibration" have been designed for
health analysis as they only include possibility to measure using one tri-axial sensor
(i.e. seat pad). None of these equipment feature a possibility to measure 12-axis
accelerations for comfort assessment.
2.5.3 Measurement procedure
Sensors are installed to seat surface, backrest and floor (Figure 16). The distance of
the backrest sensor from the seat surface should be about half of the height of the
backrest. A floor sensor should be as close as possible to feet and aligned with the
seat and backrest sensors.
50
Figure 16. 12-axis measurement system installed on a bus seat according to standard guidance
(Yamashita and Maeda 2003).
There are no restrictive guidelines on how long to measure. It is suggested by ISO

2631-1 that measurement should be at least 108 seconds and multiple repeats should
be made for improving statistical accuracy. All relevant information about the
environment and installation should be documented, as well as the vehicle and seat
information.
2.5.4 Variation relating to the measurement results
Normally vibration is measured in short duration (60 seconds to few minutes). One
problem with short term measurements is that the variability might be high. This means
that the measurements might not give statistically good values. Based on the short
term measurements it is also hard to even quantify how much variability there is in
certain work environments. Not much effort has been taken to quantify the variability of
vibration exposure of work machines in work environments (Newell, et al 2006). It has
been shown that some work machines tend to have different vibration levels even at
the same work environment, because of changing circumstances (e.g. operator, speed,
loads). Based on the current way of measuring it is difficult to evaluate the individual
effects of vibration to a person in a period of time. One factor that has not been much
discussed in the measurement publications is that the measurement setup has also
psychological effects on the operators. This means that it is very hard not to disturb the
operator in any way when conducting the measurements in a “real” work environment,
because at the minimum the operator knows it is being measured. Newell et al.
51
(Newell, et al 2006) reported a suspicion that an operator in their study drove
aggressively just because of the measurements. The author has also had a similar
personal experience.
2.5.5 Summary
Although ISO 2631-1 describes the directions, locations, frequency weightings,

multiplying factors and guidance to interpret the values of 12-axis measurements, it
does not give adequate technical details or references on how to build the necessary
sensor system. There is a reference only to the requirements of a 3-axis sensor pad for
health evaluation and to the general technical specifications of the measurement
equipment.
The availability of affordable and practical equipment is an essential part of research as

especially comfort research will require many measurements to have any statistical
meaning. If equipment is not available then other research areas are chosen, as is
evident based on the literature research on comfort.
2.6 Standards and guidelines

Whole-body vibration measurement procedures and analysis methods have been
standardised since the 1970’s (Table 4). The most important and widely used
standards have been ISO 2631-1 series (1974, 1985 and 1997) and the British
standard BS 6841 (1987). The work of BS 6841 has been closely linked to ISO 2631-1
work, because some of the groups working with the standards have been the same
(Griffin 1998). There are other methods and standards available as well, but they have
been more specific and local, such as German and Austrian standard VDI 2057 and
Average Absorbed Power (AAP) method used by USA military (Els 2005) (see
methods in Section 2.4.2). However, most of the national standards around the world
are based on the ISO 2631-1 methods.
The purpose of the standards has been to systemise the process of analysing and
evaluating the effects of vibration exposure. In short, the standards include the
analytical equations for calculating representative values and procedures for
performing the measurements. They also include technical detail and requirements for
the measurement equipment and installation of the sensors. They refer to sub set of
standards that define more precisely the techniques for measuring and signal
processing the vibration.
52
Table 4. Publication dates and code names of the main vibration standards.
Year Standard
1974 ISO 2631
1979 ANSI S3.18
1985 ISO 2631/1
1987 BS 6841
1997 ISO 2631-1
The standards help to systemize the research and evaluation. This has been very
successful in whole-body vibration research, because practically all publications have
used the methods proposed by the standards or at least compared the results to them.
Most of the publications have used the frequency weighted r.m.s. method described by
the ISO 2631-1 standard.
The review will only consider the newest standard versions of ISO 2631-1 and BS
6841, because they differ from the older standards significantly.
2.6.1 ISO 2631-1 (1997)
2.6.1.1 Background
ISO 2631-1, which is the only international standard for evaluating whole-body
vibration, has been updated over a 12-13 year period to the current 1997 version.
Currently an amendment is being finished, which will introduce small additions to the
standard, but will not affect the basic principles and methods (ISO/TC 8/SC 4 2008).
The standard defines the evaluation procedures for health, comfort, perception and
motion sickness (ISO 2631 1997). The standard describes a 12-axis measurement
procedure for comfort evaluation of a seated person. This includes acceleration
measurements from the backrest, seat and feet. The seat measurements should
include also rotational axes. The health part is focussed on a seated posture and
effects to the lumbar spine, as the standard acknowledges that this situation is the only
one where enough research has done. The standard is used widely around the world. It
is in frequent use mostly in western countries and especially in Europe, but also in
many other areas (e.g. South Africa).
53
2.6.1.2 Content of the standard
The standard regards comfort as a subjective opinion caused by a multi-axis vibration

(Figure 17).
Figure 17. Axes and locations where vibration should be measured for analysing discomfort of a
seated person (ISO 1997).
The main method for analysing all effects from vibration exposure is the frequency
weighted root mean square (r.m.s.) (Equation 4). Before the r.m.s. values can be
calculated, the acceleration data needs to be frequency weighted (i.e. filtered). The
purpose of weighting the acceleration data is to model the frequency response of the
human. The standard gives analytical equations for designing the frequency weighting
filters, thus there is an option to use different filtering schemes (i.e. digital or analogue).
There are weighting curves defined for all axes, measurement locations and
applications. The standard determines also multiplying factors for each of the axes to
compensate for the different vibration effects at different locations / directions (Table 5).
54
Table 5. Frequency weightings and multiplying factors defined in ISO 2631-1 for analysing 12-
axis measurements for a seat person for evaluating discomfort.
Multiplying
Location Direction Weighting
factor
Backrest Fore-and-aft Wc 0.80
Backrest Lateral Wd 0.50
Backrest Vertical Wd 0.40
Seat Fore-and-aft Wd 1.00
Seat Lateral Wd 1.00
Seat Vertical Wk 1.00
Seat Roll We 0.63
Seat Pitch We 0.40
Seat Yaw We 0.20
Floor Fore-and-aft Wk 0.25
Floor Lateral Wk 0.25
Floor Vertical Wk 0.40
There are three principal weigthing curves, that are used for evaluation motion
sickness, health, discomfort and perception of whole-body vibration (Figure 18).
Additionally there are three weighting curves (Figure 19) that are used in special cases,
such as in 12-axis comfort analysis. For the full 12-axis discomfort evaluation four
different weighting curves are needed.
Figure 18. Principal weighting curves of ISO 2631-1 for vertical and horizontal axes for
evaluating comfort. Wk is used for vertical axis from seat and all floor translational axes. Wd is
used for seat and backrest horizontal axes (ISO 2631-1 1997).
55
Figure 19. Weighting curves of ISO 2631-1 for rotational (We) and backrest fore-and-aft axis
(Wc) (ISO 2631-1 1997).
A crest factor for each axis should be calculated. The crest factor is defined as the ratio
between the highest frequency weighted acceleration peak divided by the frequency
weighted r.m.s. value. If the ratio is over 9, then the standard requires the use of
additional methods, because the r.m.s. value will most likely underestimate the
evaluation. The additional methods are “running r.m.s.”, which is calculated the same
way as r.m.s. except using a small integration window (e.g. one second), and the
vibration dose value (VDV), which is a fourth power dose method (Equation 6). The
reason that the running r.m.s. value is calculated is to determine a maximum transient
vibration value (MTVV), which simply is the highest single running r.m.s. value of the
measurement period for each direction. If the crest factor limit is exceeded, then either
MTVV or VDV values should be compared to the r.m.s. values using the following
criteria:
MTVV
= 1 .5 (11)
aw
VDV
= 1.75 (12)
aw ⋅T 1 / 4
where aw is a frequency weighted r.m.s. value and T is time of measurement period (s).
If either of the values in Equations 11 or 12 is exceeded (depending on which one is

used) then both the r.m.s. and additional evaluation values should be reported. No
additional information on how to use the running r.m.s., VDV or MTVV values or which
one of the methods to choose is given in the standard.
56
Next step is to combine the frequency weighted r.m.s. values using a vector sum
equation for all measurement locations. This produces a single value called “point
vibration total value” (PVTV) for each location that is measured (for the seat the
translational and rotational values are calculated separately). Each r.m.s. value should
be multiplied with a multiplying factor depending on the posture, vibration direction and
location, when values are vector summed. If any direction in any location is less than
25 % of the highest value in the same location, then that value can be excluded from
the vector sum (i.e. neglected from the evaluation).
When the PVTV for each location has been produced an “overall vibration total value”
(OVTV) can be calculated by r.s.s. for all locations. If any PVTV in any location is less
than 25 % of the highest PVTV in another location then that value can be excluded as
well. OVTV can then be compared to other such values from different measurements
and/or additionally to subjective comfort levels described in the standard’s Annex C
(Table 6).
PVTV (apvtv) of translational or rotational axes from a location is calculated:
a pvtv = k x2 awx
2
+ k y2 awy
2
+ k z2 awz
2
(13)
where kx, ky and kz are the multiplying factors for the axes x, y and z (or roll, pitch and
yaw) and awx, awy and awz are the frequency weighted r.m.s. values of x, y and z (or roll,
pitch and yaw) axes. OVTV (av) is calculated by combining necessary PVTVs:
av = ∑aj
2
j (14)
where aj is a point vibration total value of location j and j is either seat, backrest or floor.
The report of the analysis should include the frequency weighted r.m.s. values (with
multiplying factors) for each axis, point vibration total values (PVTV) for all locations
and the overall vibration total value (OVTV). If the crest factor limit of 9 is exceeded,
then also the crest factor and any additional methods and their results should be
reported as well. Also measurement duration, frequency spectra and any conditions
that might be important should be reported in any case. OVTV can be compared to the
subjective response scale given in the standard’s annex C (Table 6), although it is
given only as an approximate indication.
57
Table 6. Approximate indications of likely reactions to various magnitudes of overall
vibration total values in public transport (ISO 2631-1 1997).
Magnitude of overall vibration total value Discomfort response

2
Less than 0.315 m/s Not uncomfortable
2 2
0.315 m/s to 0.63 m/s A little uncomfortable
2 2
0.5 m/s to 1.0 m/s Fairly uncomfortable
2 2
0.8 m/s to 1.6 m/s Uncomfortable
2 2
1.25 m/s to 2.5 m/s Very uncomfortable
2
Greater than 2.0 m/s Extremely uncomfortable
Figure 20 shows the standardised measurement and analysis procedure for seated
persons of the standard, which was described above.
58
Figure 20. Measurement and analysis procedure of ISO 2631-1 standard for evaluating comfort
from whole-body vibration.
2.6.1.3 Interpretation of ISO 2631-1
The standard implies that vibration causes discomfort and that the mechanism of it is
complex. This can also be said of the procedure that the standard describes for
evaluating it. Although it is logical that the r.m.s. method cannot be used in every
situation, the guidance on when it is appropriate to use it is confusing. Griffin has
explained this for the health evaluation procedure (Griffin 1998), which has the same
basic problem.
59
As this study is focussed on comfort, the information relating to it is evaluated. First of
all, compared to 1985 version (ISO 1985), the newest version (ISO 2631-1 1997) is
more careful of what to conclude from the analysis of comfort:
“For simplicity, the dependency on exposure duration of the various effects on people
had been assumed in ISO 2631-1:1985 to be the same for the different effects (health,
working proficiency and comfort). This concept was not supported by research results
in the laboratory and consequently has been removed. New approaches are outlined in
the annexes. Exposure boundaries or limits are not included and the concept of
“fatigue-decreased proficiency” due to vibration exposure has been deleted.”
Thus it shows that new research has concluded the evaluation of comfort to be even
more complicated than previously understood. The standard then continues to explain
the causes of vibration exposure:
“Whole-body vibration may cause sensations (e.g. discomfort or annoyance), influence

human performance capability or present a health and safety risk (e.g. pathological
damage or physiological change).”
This shows that discomfort and problems relating to it are important and
acknowledged, even though the methods for analysing them are still inadequate.
So far the standard has been consistent with its statements. However when reading
further the problems of the standard and the complexity of the subject is more evident.
Just after introducing the basic measurement guidance and the r.m.s. method the
standard narrows the usage of it by stating that crest factor exceeding 9 will most likely
mean that r.m.s. is underestimating the effects of vibration. In a note the standard even
further undermines the r.m.s. method:
“For certain types of vibrations, especially those containing occasional shocks, the
basic evaluation method may underestimate the severity with respect to discomfort
even when the crest factor is not greater than 9.“
When the crest factor limit is exceeded the standard implies that the basic evaluation
method (i.e. r.m.s.) most likely underestimates the effects and refers to the additional
methods (i.e. VDV, running r.m.s. and MTVV). However, the r.m.s. method is still
considered the main method and should be reported in any case. There is no guidance
on how to use the additional methods for comfort evaluation or comparison with the
other values. They are only used for evaluating the usability of the r.m.s. method.
60
The clauses for comfort evaluation are also inconsistent, because they confuse the
usage of the method. First of all the application paragraph states that:
“For the comfort of seated persons this clause applies to periodic, random and
transient vibration in the frequency range 0.5 Hz to 80 Hz which occurs in all six axes
on the seat pan (three translational: x-axis, y-axis and z-axis and three rotational: rx-
axis, ry-axis and rz-axis). It also applies to the three translational axes (x, y and z) at the
backrest and feet of seated persons...“.
It further goes on explaining comfort in next clause by stating:
“The weighted r.m.s. acceleration (see clause 6) shall be determined for each axis of
translational vibration (x-, y- and z-axes) at the surface which supports the person.”
This statement only describes translational axes, not rotational. It might give the wrong
impression, especially if the previous clause has not been read well, that only
translational axes are necessary to be measured. In the main text the standard only
gives multiplying factors for the three translational axes from the seat. However note 3
in the clause informs that:
“In some environments, the comfort of a seated person may be affected by rotational
vibration on the seat, by vibration of the backrest or by vibration at the feet. Vibration at
these positions may be assessed using the following frequency weightings…”
And then the multiplying factors and frequency weighting curves are given. After
reading the three paragraphs stated above it seems that it leaves the evaluation of how
to measure comfort to the user, because there is no clear guidance on if the full 12-axis
measurements are needed. The standard does not state that you have to measure all
axes. It gives that possibility, but emphasises only the translational axes on the seat
and refers to other axes if they affect the comfort evaluation. Because there is no
guidance on when to measure them, it is hard for the user to conclude what axes
should be included. Without measuring the twelve axes it might be difficult to conclude
what axes actually cause discomfort, especially when the standard does not give any
guidance on what environments this should be considered. Additionally the size of the
calculated OVTV value is related to number of included axes, thus less axes will mean
less discomfort.
The standard gives a possibility to use the translational axes measured from the seat to
estimate the effects of backrest using 1.4 multiplying factors for horizontal axes. There
is no reference to any study that can confirm the validity of the multiplying factors and
the guidance is given in a small note, which can be easily missed.
61
For combining the values of the axes the standard states that:
“Where the comfort is affected by vibrations in more than one point an overall vibration
total value can be determined from the root-sum-of-squares of the point vibration total
values (e.g. translation on the seat and at the back and feet).”
Again the standard leaves an option to either combine the values or leave them
separate. It also allows excluding one or more axes if certain requirements are met:
“If the weighted value determined in any axis (or rotational direction) is less than 25 %
of the maximum value determined at the same point but in another axis (or rotational
direction) it can be excluded. Similarly, if the point vibration total value in one point is
less than 25 % of the point vibration total value which is maximum, it can be excluded.”
Thus, in theory only one axis could be enough for comfort evaluation and assessment
(e.g. when a large shock for any axis occurs). In practice this in not the case, at least
not with mobile work machines, as the dominant axes change depending on the work
phase. In worst case this does not lead to any significant differences between values
and goes well in the limits of measurement uncertainty (Tyler and Darlington 2004), but
it still confuses the user and alters the results. Although the guidance for doing that is
given in a note, it is hard to conclude if this procedure is recommended or not. There
are no studies or references which confirm if exclusion of any values is valid.
The standard refers to annex C for more information on evaluating comfort and
perception. However for comfort the annex does not give any new information. It
actually limits the usability even more by stating that comfort feeling changes based on
the circumstances and many other factors. Even though the standard has twice stated
that there is no conclusive evidence of any time dependency of vibration on comfort, it
gives an equation to evaluate it. For methods for assessment when the crest factor limit
is exceeded the annex refers back to the main part of the standard. The annex gives
also the subjective response scales for evaluating the results. However the scales
overlap, thus human reaction to comfort cannot be exclusively evaluated. There has
been no research or solutions to solve this problem (Wyllie and Griffin 2007).
Criticism of the standard has increased over the years, because many studies have
found inconsistencies regarding the frequency weighting curves and averaging
methods (Griffin 1998). The frequency weighting curves have been based on the
perception studies. There are several problems with them as the curves have been
extrapolated based on only a few frequencies and linear response has been assumed,
which is not consistent with some studies. For evaluating comfort the problem is also
62
that the weightings are statistical, thus they give no individualised effect (i.e. no
personal information is regarded in the analysis). There is also an indication that W k,
the vertical frequency weighting curve, underestimates subjective feeling of lower
frequencies (Mansfield and Maeda 2005b).
2.6.1.4 Important changes to earlier versions
The evolution of ISO 2631 has been significant from the early version to the current.
Earlier studies have shown that first draft did not include a concept of forming a single
numerical value to determine effects from vibration exposure (Miwa and Yonekawa
1974). The biggest change has been from the 1985 version to the current 1997
version. More specific frequency weightings and evaluation methods were
implemented and new definitions applied. More restrictions to the usage of the
standard were also applied, based on the new information. Originally it was considered
only to evaluate the worst frequency component of the worst axis (ISO 1974 and 1978).
Later it was changed to taking the r.s.s. for the 1985 version. The main method in the
early standards was ‘single-axis 1/3-octaveband analysis’, but additionally ‘single-axis
frequency weighted r.m.s. value’ and ‘the sum of vector frequency weighted r.m.s.
value’ was proposed as secondary methods (Wikstroem, et al 1991). The methods in
the early versions of the standard were not considered valid or reliable (Griffin 1978).
2.6.2 BS 6841 (1987)
2.6.2.1 History
Although the UK voted against the first version of ISO 2631-1 standard (Griffin 1998),
they produced their own standard with very similar content. The standard is broader in
defining its scope and resembles more ISO 2631-1 (1997) than the earlier versions.
One cause for publishing BS 6841 was the growing interest towards vibration problems
in industry and the slowness of the international forum to produce a new ISO standard
(Griffin 1998). The role of BS 6841 has been significant since its introduction and for
the creation of ISO 2631-1 (1997).
2.6.2.2 Content of the standard
The basic method in BS 6841 is also the r.m.s. method (Equation 4). Frequency
weightings are applied in the same way as in ISO 2631-1, except for vertical direction
where weighting curve Wb is used instead of Wk. For using the full 12-axis
measurement procedure for comfort evaluation BS 6841 has the same multiplying
63
factors as in ISO 2631-1. The crest factor is calculated exactly as in ISO 2631-1,
although BS 6841 uses a lower crest factor limit (6) when referring to additional
methods.
With the exception of the crest factor limit the rest of the process is almost identical to
ISO 2631-1 for calculating PVTVs and OVTV. The only exception is that BS 6841 does
not exclude any PVTVs like ISO 2631-1 does if 25 % margin is exceeded. Only when
calculating PVTV, then BS 6841 has the same guidance to exclude axes based on the
25 % margin.
If the crest factor limit is exceeded, then the analysis procedure changes from the one
in ISO 2631-1. The standard recommends of using either root mean quad (r.m.q) or
VDV as an additional method. R.m.q and VDV are both fourth power methods, so
basically either one can be used (but r.m.q. is implied). VDV is recommended to be
used if two or more events of different duration are needed to be analysed together.
After calculating either the r.m.q. or VDV values the standard does not refer back to
usage of the r.m.s. values, thus the values from r.m.q. or VDV calculations should be
reported (Figure 21).
64
Figure 21. Measurement and analysis procedure of BS 6841 for evaluating comfort of seated
persons from whole-body vibration.
2.6.2.3 Interpretation of BS 6841
The discomfort is defined in the standard as “sensations arising directly from the
vibration”. It defines a method for comparing the effects of multi-directional vibrations,
but reminds that the effects are related to specific environments. In a note about
human variability the standard states that:
“People differ in their response to vibration and individuals may have different reactions
at different times. Thus two motions, which are assessed as equally severe by the
recommended evaluation procedures, may have noticeably different effects.”
The statement indicates that the methodology in use is derived from statistical studies
that have averaged the results of many people, which can contain larger variations,
thus this method might give wrong results if a single person is considered.
65
An interesting note is presented about the frequency weighting:
“Weightings for rotational vibration are therefore only considered necessary below 10
Hz. However the weightings are given for higher frequencies for consistency.”
There is an indication that rotational vibrations above 10 Hz are perceived as

translational, but the analysis method still uses the rotational axes (i.e. frequency
weightings) above 10 Hz. The standard implies that defining the crest factor is
important when evaluating comfort and it should be below 6 (or even lower) for more
accurate analysis. The standard refers to alternative methods in the appendix C if crest
factor of 6 is exceeded.
As ISO 2631-1 also BS 6841 guides to exclude any direction that has 25 % lower
value, but only for calculating PVTVs:
“if a weighted value determined in any axis is less than 25 % of the maximum value
determined at the same point, but in another axis it need not to be included in the
above calculations.”
For the measurement procedure the standard states the same 12-axis procedure as in
ISO 2631-1 (1997) for seated postures. For standing and recumbent postures the
method is also the same 3-axis procedure from the seat, than in ISO 2631. Frequency
weighting curves are given to all axes with multiplying factors.
In an appendix C, where discomfort is further explained the standard states the same
things about variability in humans and environments. For vibration containing a high
crest factor (>6) the standard suggests:
“For these situations the root-mean-quad value will be more likely to predict accurately
the relative discomfort of motions of similar duration than r.m.s. value.”
It further goes on to give even another possibility:
“If vibration exposures consist of periods of high and low vibration throughout variable
periods of time, the vibration dose value may be determined as described in the
appendix A.”
The r.m.q. and VDV are fourth power methods, which are given as an option, although
no other information about the usability of the methods is given. The standard then
gives a guideline for comparing the r.m.s. values for subjective opinion. In a note the
standard states:
66
“The procedure specified in this standard is based upon laboratory research and field
surveys.”
BS 6841 standard also emphasises only three translational axes from the seat and
states in a note the same thing as ISO 2631-1:
“In some environments the comfort of a seated person is affected by rotational vibration
on the seat, by vibration of the backrest or by vibration at feet.”
After that it gives in smaller print the rest of the multiplying factors and frequency
weightings for other axes.
2.6.3 Comparison of ISO 2631-1 and BS 6841 standards
A quick look at the standards makes them seem very similar. The primary method in
both is the calculation of the frequency weighted r.m.s. values for each of the twelve
axes. The multiplying factors are the same as using the crest factor to evaluate
applicability of the r.m.s. method. There are differences in the vertical frequency
weighting curve Wb of BS 6841 to Wk ISO 2631-1, but they are marginal (other
frequency weightings are exactly the same). BS 6841 uses lower crest factor limit than
ISO 2631-1 for recommending using of additional methods (the same as ISO 2631-
1:1985 uses), but this can also be stated as a small difference (although the usability of
the r.m.s. method itself is greatly lowered).
The biggest differences between the standards are when the crest factor limit is
exceeded and additional methods are required. ISO 2631-1 refers to the running r.m.s.,
MTVV and VDV methods, but actually only instructs to use them to find out if the r.m.s.
method is correct (there is no indication of how to interpret the values themselves).
Even though the r.m.s. method is said to underestimate the vibration ISO 2631-1
instructs to always report it as well. BS 6841 on the other hand refers to the r.m.q. or
VDV methods as an option, but only indicates that the methods will “more likely” predict
the relative discomfort. Both standards give no guidance on how to interpret to results
of the additional methods, thus their usage does not give any useful information.
Especially for the r.m.q. method where there is no additional information in the
standard. Only a few publications have used the r.m.q. method since its existence
(Boileau, et al 1989, Fairley 1995, Monsees, et al 1988), thus not enough information
on usability and its relation to comfort exists. The standard has only additional
information about VDV relating to health.
67
In general BS 6841 is simpler and more understandable than ISO 2631-1, as it does
not have that many notes that only confuse the reader. The problem with both
standards is that they rely on a method that underestimates even relatively low shocks
and spikes, thus they have to spend time on describing limitations for the usability and
introduce additional methods, which have insufficient guidance and no greater scientific
proof than the default method.
Another issue is that there is no time dependence for comfort. Basically one minute of
vibration exposure can be concluded to be as comfortable as 10 hour exposure if the
r.m.s. or r.m.q. method is used. Only the usage of VDV incorporates time dependency,
but there is no conclusive knowledge on how duration affects comfort feeling. Certainly
there are no tables for assessing discomfort for the VDV values.
Both standards also give the possibility of not measuring rotational axes or backrest
and floor. They only loosely state that if other axes or locations have an effect on
comfort then they should be considered. The guidance on how to assess if they are
needed is not given, thus it is easy to exclude them and just use the current
commercial seat pad with three axes for comfort measurements. The 12-axis
measurement procedure of the standard is optional, not mandatory, and is only given
as an informative reference as there is no real evidence of the method to be able to
estimate comfort in practice. If no better guidance is given, then it is assumed that all of
the twelve axes should be measured for the best accuracy.
The methods in both standards are based on the knowledge gathered from studies that
have used single-axis or dual-axis vibration exposure. There is a reason to assume
that multi-axis vibration exposure will have a different effect than combinations of
single-axis exposures. There is a difference in biodynamic response of multi-axis
versus single-axis (Holmlund and Lundström 2001). This could indicate that also
comfort opinion will change in multi-axis environments.
It is clear that the possibility of different variations of procedure, limitations of the basic
method, lack of guidance of additional methods, small printed notes and scattered
information make the standards very hard to use in a systematic manner (or even
properly). Depending on the user the procedure can show different results. This has
been concluded already for the health evaluation procedure (Griffin 1998), but it is
more evident for both standards in comfort evaluation procedure.
For any method or procedure to be practically used it should not have a possibility for
differing interpretation. It should clearly define the methods and limitations to every
68
situation. It seems that the only practical procedure to measure and analyse comfort
based on the standard is to always measure the 12-axis vibration, calculate frequency
weighted r.m.s. values, calculate the crest factor values, and calculate additional
methods and report all of them separately for each axis.
2.6.4 Additional standards
2.6.4.1 ISO 10326
The standard specifies: “basic requirements for the laboratory testing of vibration
transmission through a vehicle seat to the occupant”. The standard has a description
for a 3-axis seat pad (Figure 22) and installation locations for floor, backrest and seat
surface used in whole-body vibration measurements. The measures for the 3-axis seat
pad can be considered as a basis for designing the 6-axis seat pad’s mechanical
housing. The standard also introduces the SEAT value and equation.
Figure 22. Seat pad sensor measures according to ISO 10326 (1992).
2.6.4.2 ISO 8041
ISO 8041 defines the test procedure, tolerances and technical information for
equipment used in whole-body vibration measurements. There are no direct
specifications concerning 12-axis equipment or 6-axis seat pad sensor, but most of the
specification intended for the 3-axis equipment can be used directly, such as
tolerances for the frequency weightings, r.m.s. values, and so on. The requirements
are not considered here in detail, but they are used when developing the equipment in
Chapter 4.
2.6.5 European legislation
The European Union accepted and ratified a new directive (2002), which mandates
minimum health and safety requirements regarding the exposure of workers to the risks
69
arising from physical agents (vibration). The directive sets limit values in form of r.m.s.
and VDV values that dictates the vibration a worker can be exposed to during an eight-
hour work day. The directive sets certain requirements that the employer must follow to
reduce the vibration affecting the worker.
The directive does not refer to comfort evaluation or to any problems regarding comfort
and working ability. Because of that, the directive is not related to this thesis directly,
but it is still useful to understand the basics of the directive and how to relate the results
based on the directive to comfort evaluation.
There are two limit values that are used: 1) daily exposure limit value and 2) daily
action value. The daily exposure limit value (1.15 m/s2) can never be exceeded in the
standard eight-hour working day. The daily action value (0.5 m/s2) sets a limit as to
when an employer should start a program to reduce the vibration level. The directive
lists basic measures that the employer should do when the vibration levels are too
high. Although the directive sets limit values to act upon, the directive also instructs and
expects the employer to reduce and evaluate the vibration in any case. From the
measurements only the highest value of the three axes (x, y and z) from the seat is
used to assess the vibration exposure.
2.6.6 Guidelines for assessing discomfort from vibration
There is no practical guidance for comfort assessment as there is for health (EN 14253
2003). The guidance does not have any information on assessing comfort from
vibration. It only refers to comfort of seat and other ergonomic factors, not on how to
evaluate vibration related to comfort using the standardised method.
2.6.7 Summary
Based on the review of the standards and their guidance on measuring and evaluating
the health of seated persons, it is evident that there are difficulties in using them. The
author has not been able to find any previous publications, which have critically
evaluated the comfort evaluation procedure of ISO 2631-1 and the procedure for
validating it using the full method and field-like stimuli. Some field studies have been
made, but the results have not been confirmed in a laboratory environment. Also rarely
have all twelve axes been measured.
For any method or procedure to be practical it should not have a possibility for differing
interpretation. It should clearly define the methods and limitations to every situation.
Based on the review of the standard and publications it is clear that the guidance gives
70
a possibility for large differences in the final results even though it has been properly
followed.
The comfort evaluation method has not been validated as a whole. The method is
based on separate studies, which have all used only one or two axes simultaneously or
a limited setup (Maeda and Mansfield 2006a). There are no known publications or
research, which have validated the overall procedure in a multi-axis environment. It is
evident that research on validating the whole method is needed and this includes more
measurements in field and in the laboratory. Additionally even though it has been
shown to work reasonably well for stationary vibrations (e.g. sinusoidal), the
applicability in the field with and without high shocks has been questioned (Wikstroem,
et al 1991).
Even if the standard method is valid the problem of confusing guidance and lack of
practical instrumentation still hinders its usage. More practical and affordable
equipment should be developed to enable wider usage of the standard. The standards
describe the equations, sensor directions and locations, frequency weightings,
multiplying factors and general guidance to measure and interpret the values for
analysing discomfort. They do not give any technical details or references on how to
build the necessary instrumentation, nor detailed information on the installation
procedure.
2.7 Conclusions
Although there is evidence of negative effects of vibration to health and comfort, the
details of the reasons are still mainly unknown. The modern belief is that vibration is
just one of the many factors that contribute to comfort and health problems (Hostens
2004, Hoy, et al 2005). Still health and comfort problems seem to appear more likely
with professional drivers than any other groups (Agius, et al 1988, Boshuizen, et al
1990, Bovenzi and Betta 1994).
This review has shown the effects and importance of understanding vibration. It is
shown that the link between vibration characteristics and discomfort is not yet
completely understood and is complex. Based on the review there is a need to improve
understanding of the standard method for predicting discomfort from whole-body
vibration. The process of validating the standard method requires the development of
more accessible technical equipment, which allows research in the field and laboratory
and gives possibility to validate the standard procedure and to further improve it.
71
3 Chapter 3 – Planning and selecting
general methods for validating and
improving the ISO 2631-1 method
3.1 Introduction
Based on the literature review it was evident that validation of the ISO 2631-1 (1997)
standard method is necessary in a multi-axis environment and using field-like stimuli.
Most of the studies have been conducted in a laboratory using single or dual-axis
stimuli and using sinusoidal or narrow-band random stationary excitations. Typically
exposures have been short (i.e. a few seconds) and judgments have been acquired
using methods more suitable for the laboratory. The studies which have been
conducted in the field, have shown conflicting results, have rarely used all twelve axes
and have mainly used test tracks with shock type stimuli.
To introduce new results and validation of the standard it is necessary to conduct trials
which are closer to real world circumstances. Thus methods, which can be used in field
environments, are needed for acquiring and analysing judgments. Methods used in the
previous studies are not practical in the field or for long exposures as they only include
one judgment per stimulus and might need a reference stimulus (Maeda 2005). The
judgment of time variant vibration (i.e. random vibration) has been shown to be
problematic using traditional methods (Maeda 2005).
Even though field trials will include more variability, it is important to validate results of
laboratory studies in the field, otherwise there is no link to practicality. There is a clear
need to improve testing of seat comfort in the field and methodology to analyse the
results (Kolich 2008).
3.2 Research plan and goals of the thesis

The purpose of the thesis work is to validate the applicability of the current ISO 2631-1
standard in laboratory and field environments using multi-axis field-like stimuli.
Depending on the results the second goal is to improve the predictability of the
discomfort from whole-body vibration either by optimising the current standard method
or by suggesting a completely new method. The first priority is to optimise the current
standard method, as this would comply better with the current equipment and
guidance.
72
To limit factors and scope of the work a seated passenger exposed to whole-body
vibration while not conducting any specific task is regarded as the study case. This
allows more detailed study of the relative effects of axes and correlation of discomfort
and vibration characteristics.
Thus based on the problems with the current standard guidance and its usability it
becomes clear that the task for generating more research in this area, is to suggest
more logical evaluation procedure, develop more practical equipment and improve
guidance for comfort assessment. Based on the conclusion the following steps must be
taken, which will be the goals of this thesis:
• Develop practical equipment for measuring 12-axis vibration in the field and
laboratory;
• Validate the developed equipment using laboratory experiments;
• Measure and evaluate 12-axis vibration from the field from different machines;
• Evaluate correlation of discomfort to vibration using six axis test bench in a

laboratory;
• Optimise ISO 2631-1 discomfort model based on laboratory results;
• Evaluate correlation of discomfort to vibration using a real vehicle in the field;
• Optimise ISO 2631-1 discomfort model based on field results;
• Suggest more practical guidance based on the results.
3.2.1 Goals for developing and validating equipment
The literature review showed that commercial equipment for measuring all twelve axes
of vibration are expensive and difficult to obtain. For a researcher to build their own
measurement system requires understanding of electronics, sensors, signal processing
and software programming. A simple and inexpensive, yet accurate, system for
measuring all twelve axes was considered necessary. This will include a six-axis seat
sensor and two tri-axis sensors for backrest and floor. Also software for calculating
translational and rotational axes will be developed. The equipment should allow long
measurement periods (several minutes) of raw acceleration data. The equipment is
developed to be commercial quality, thus allowing other researchers to purchase the
equipment and software. Two experiments will be conducted to validate the equipment
in a multi-axis test bench and in the field.
73
3.2.2 Goals of the 12-axis field measurements
Based on the literature review there were only a few 12-axis measurements available
publicly (Griffin 1990, Maeda 2004). It was concluded necessary to conduct additional
field measurements from typical work environments to have more confidence in the
analysis. The goal of the field measurements is to obtain enough data to analyse
relative contribution of the twelve axes in normal work environments. The results are
compared to other publications and concluded how many axes have practical
contribution to the overall vibration total value, does contribution change for different
machines and environments, and to provide guidance which axes are sufficient for a
full discomfort evaluation and the number of axes which can be excluded based on the
‘25% contribution’ criteria.
3.2.3 Goals of the trials
Studies which have claimed or introduced a new method to predict discomfort or

health, have been based on either laboratory or field trials. Laboratory studies have
normally a very narrow scope of stimuli and an artificial environment. On the other
hand assumptions made in field studies can be inappropriate and error margins too
large (Griffin 2007). It is necessary to incorporate both research environments for
testing and evaluating methods in laboratory and validating them in the field, thus this
thesis uses both approaches.
3.2.3.1 Laboratory trial
The goal of the laboratory trial is to validate the standard method for predicting
discomfort from whole-body vibration and to analyse how frequency weighting,
multiplying factors and averaging methods affect the results in a multi-axis environment
using non-stationary stimuli. The trial will conclude how many and which axes are
necessary for predicting discomfort and how correlation between discomfort and
vibration varies amongst subjects. The trial results will also be used to try to optimise
the standard method based on the analyses of effects of different parts of the methods.
3.2.3.2 Field trial
The goal of the field trial is to validate and test the standard and improved method in
practice. The field trial will introduce several variables, which are not present in a
laboratory. It is important to test the method in the field as it shows how practical the
standard method is. Several subjects will be measured simultaneously as no
74
permanent test track (i.e. highly repeatable vibration exposure characteristics) is used.
This requires a simple method and equipment to acquire judgment data.
3.3 Selecting general methods for conducting the research
3.3.1 Calculating vibration magnitudes
Vibration magnitudes for each axis are calculated based on the measured acceleration
data in m/s2. The calculations are performed using the guidance from the main
standards (ISO 2631-1 and BS 6841). Because of the nature of the thesis the
magnitudes are calculated using several different approaches, where some
magnitudes are calculated without applying parts of the standard method. Table 7
summarises different approaches of calculating the vibration magnitudes that is used in
this thesis.
Table 7. Different approaches used in the thesis to calculate vibration magnitudes for each axis.
Frequency weighting and multiplying factors are applied based on the ISO 2631-1 standard.
Averaging methods are applied based on the ISO 2631-1 and BS 6841 standards. Point and
overall vibration total values are calculated for all setups.
Frequency Multiplying
Setup Averaging method
weighting factors
1 Yes r.m.s. Yes
2 Yes r.m.q. Yes
3 Yes r.m.s. No
4 Yes r.m.q. No
5 No r.m.s. No
6 No r.m.q. No
3.3.2 Acquiring and calculating subjective judgments
Normally trials include a short stimulus and an overall judgment is made for each
stimulus. This is possible if the stimuli are relatively short (i.e. less than 15 seconds).
However for longer stimuli the overall judgment will most likely be unreliable as
subjects need to memorise the feeling during the exposure (Kuwano and Namba
1985). Thus a method which can be used for more real-time judgment is a necessary
requirement.
Based on the requirements and aims of the analyses it was decided to use the
continuous judgment method of line length to acquire judgment data simultaneously
with vibration data (see Chapter 2 Section 2.3.2.6 for details of the method).
75
3.3.3 Statistical analysis
Because of the stimuli types, measurement environments and chosen judgment

method the correlation needs to be calculated and analysed using a ranking method,
as normal distribution cannot be assumed. A Spearman rho (r and r2) is selected as the
primary method for calculating correlation, because it is a widely used method and
simple to calculate. Additionally Kendall’s tau (t) is calculated for the most important
results for giving a better insight for the understanding of the relationship and
confidence on the results. In most cases both methods will lead to the same
conclusions (Crichton 1999).
A Wilcoxon signed-rank test is used to evaluate statistical differences amongst two sets
of matched data (e.g. differences of correlations for each subject). A Friedman test is
used to evaluate if the number of dependent results have any statistical differences. A
Friedman test can be applied to multiple data sets. A Mann-Whitney U-test is
conducted when the two data sets are independent. The significance of the results is
evaluated based on the p-value (two-tail) and null hypotheses are introduced for each
case.
3.4 Conclusions
Based on the literature review in Chapter 2 it was found that ISO 2631-1 needs to be
validated and possibly improved for comfort evaluation in the field. A number of
experiments, trials and field measurements were planned, which require robust and
simple methods to analyse the results. Several approaches are used to calculate
vibration magnitudes and a number of statistical methods are used for analysing
correlation between subjective judgments and objective results.
76
4 Chapter 4 - Developing and validating
equipment for measuring 12-axis
vibration in a field and laboratory
4.1 Introduction
There are only few publications, which have used the full discomfort evaluation method
of ISO 2631-1 (1997) or BS 6841 standards (Griffin 1990, Maeda 2004). This is partly
because of lack of practical equipment and the complexity of the standardised method.
In Chapter 2 there was an analysis of the current discomfort evaluation procedure of
the standard and the required work to validate it. Based on the review the development
of more practical and less expensive equipment for 12-axis measurements was
concluded. Currently there is no feasible commercial equipment to allow wider usage of
the 12-axis comfort method in practical field measurements, because the equipment is
expensive and not actively sold. Without proper equipment it is not possible to increase
the knowledge on how to evaluate discomfort in practice.
Based on the previous publications it has been found that the standard does not give
an explicit guidance on the 12-axis measurements for comfort evaluation. There is a
possibility for interpretation when analysing the measurement results. The full method
requires measurements from a seat, backrest and floor. The measurements from floor
and backrest can be done using traditional 3-axis accelerometers that record
translational accelerations, but the measurements from the seat needs also rotational
accelerations. There is a doubt how many axes are needed for assessing discomfort of
a seated person. These problems have hindered the usage of the comfort method.
4.2 Goals of the development of equipment

New measurement equipment was to be designed, developed and validated for
allowing practical 12-axis field measurements. The purpose was to realise the
measurement instrumentation so that it is practical and simple for field measurements
and made from affordable components. The equipment will be validated in laboratory
experiments and a commercial partner will be searched for producing the equipment
for open market.
The development was focussed on a 6-axis seat pad sensor, which included detection
of both translational and rotational axes. Additionally the data acquisition and signal
77
processing hardware was designed as well as software for analysing the measurement
data.
4.3 Developing and validating measurement equipment
4.3.1 Components and methods for acquiring and detecting translational

and rotational axes
4.3.1.1 Acceleration MEMS sensors
The sensor system was chosen to be based on MEMS technology. The advantage of
the MEMS sensor (i.e. capacitive sensor) compared to a piezoelectric sensor is a
simpler design and needs very few extra components before digitizing the signal. The
sensitivity and noise levels are well within the technical requirements of the standard
(ISO 8041 2005). MEMS sensors have typically voltage in – voltage out operation. In
practice the sensors only require low noise voltage source (e.g. 5V) and proper anti-
aliasing components, and an A/D sampling circuit. The sensors also detect gravity, not
like piezoelectric sensors, so in practice the measurements can be made up to a DC
(i.e. 0 Hz) level. The challenge is to minimise the noise levels and to optimise the
sensor’s whole dynamic range, which is less than with piezoelectric sensors. For
whole-body vibration measurements this can be achieved if the measurement accuracy
is specified to ±0.01 m/s2, which is well enough for practical field measurements.
Table 8 shows typical technical specifications of a MEMS acceleration sensor. Range

is typically from 1.5 to 8 g, which is directly related to the resolution of the sensitivity of
the sensor. For measuring whole-body vibration from a seat a 2 g range is sufficient.
For a floor a higher range is suggested.
78
Table 8. Technical specifications of the MEMS sensor selected for the 12-axis measurement
system.
Parameters Units Values
Range g 1.5 to 6.0
Sensitivity mV/g 560
Noise µV / Hz 175 typical
Bandwidth Hz 0-3300 (vertical), 0-1700 (horizontal)
Cross-axis Sensitivity % ±2.0 typical
Operating temperature C -40 to +85
4.3.1.2 Data acquisition
Any analog signal that is digitized will require an analog-to-digital (A/D) conversion. The
resolution of the conversion is based on the A/D characteristics (i.e. number of bits it
can use to digitize the analog signal). For an example when a 24 bit A/D sampling
resolution is used, the maximum theoretical sensitivity of the measurement device for a
certain voltage span can be calculated. The maximum 24 bit sampling resolution (Rmax)
and maximum theoretical sensitivity (Smax) using a voltage span of 5 V is:
5V
Rmax = 24
= 2.98 ⋅ 10 − 7 V = 0.000298mV (15)
2
If an analog acceleration sensor, with a sensitivity of 560 mV/g, is used in conjunction

with the A/D converter, the maximum sensitivity (i.e. resolution) of the measurement
system will be:
0.000298mv
S max = = 5.32 ⋅ 10 −7 g = 5.32 ⋅ 10 −6 m / s 2 (16)
560mv / g
Thus the better resolution of the A/D conversion the greater theoretical accuracy it is
possible to achieve.
4.3.1.3 Accuracy of an acceleration signal
The practical sensitivity of any system is affected by the noise characteristics of all
components and the sensitivity of the accelerometer. There are many noise sources
that affect the measurement results, but the most important ones are the noise
characteristics of the sensor and the A/D component. They will affect the accuracy of
the analog-to-digital conversion.
79
Based on the result of Equation 16 a conclusion can be made that the noise and
sensitivity characteristics of the accelerometer, its circuit board and cabling are the
limiting factors. There are also noise sources from the environment affecting the
accuracy of the system. So even though all the necessary information about the MEMS
sensor and other electronic components are given in the technical specifications, in
practice the whole system (including the cabling) has to be tested and validated on a
test bench.
4.3.1.4 Method for calculating rotational accelerations
For measuring rotational components of vibration a design having three sensors in

three corners of the plate (A0, A1 and A2) for calculating rotational axes, and a single 3-
axis sensor in the middle (A4) for translational axes is selected (Figure 23). Calculated
rotational accelerations are based on the relationships of the sensors’ directions and
the distances between them (Equations 17-19). The additional 3-axis sensor in the
middle is redundant, as all six axes could be calculated based on the corner sensors.
However, for maximal accuracy of the translational axes and to simplify the calculation
process, the middle sensor was used.
Dy
A2
A0
y
Dx A4 Dx/2
Dy/2
A1
Figure 23. Sensor locations and orientations for calculating 6-axis vibration using 6-axis
accelerometers at points A0, A1, A2 and A4.
80
az 2 − a z 0
Roll: aP ,rx = (17)
Dy
a z1 − a z 0
Pitch: a P ,ry = (18)
Dx
a y1 − a y 0
Yaw: a P ,rz = (19)
Dx
where azo, az1, az2, ay0, and ay1 are acceleration (m/s2) for directions x, y and z at
locations A0, A1 and A2, Dy and Dx are distances in metres between locations A0, A1 and
A2 (Figure 23).
4.3.2 Prototype I: Experiment for validating sensor configuration for

measuring 12-axis vibration
4.3.2.1 Purpose
The goal of the first prototype was to test the chosen MEMS sensor type and
configuration for measuring 6-axis vibration. An analog ±2.0g version (560 mV/g) of the
sensor was chosen.
4.3.2.2 Technical details
The 12-axis sensor system requires a measurement device for recording the signals,
which includes A/D-channels, anti-aliasing filters, processor, software and memory,
among other components (e.g. voltage regulation). The first prototype for recording the
sensor information was realised using a commercial development kit and a developed
I/O card including two 24 bit 8-channel A/D converters. A development kit from Altera
(Stratix II) was used in this study. The measurement device sampled at 260 Hz each
channel and saved the raw acceleration data to memory and the analysis process
(including the frequency weighting) was done afterwards using a Matlab environment.
Figure 24 shows the first prototype of the measurement device.
81
Data acquisition
system
3-axis sensors
6-axis sensor
Figure 24. The 6-axis sensor configuration for measuring translational and rotational axes from
seat (left) and the measurement system to acquire measurement data (right).
4.3.2.3 Validation of the accuracy of the equipment
The prototype measurements system was tested using a 6-axis test bench at Japan
NIOSH, Kawasaki, Japan (Figure 25 left). The purpose was to evaluate and validate
the developed measurement equipment and the 6-axis MEMS sensor configuration
(later called MEMS) against a 12-axis measurement system made by IMV Corporation
(later called IMV).
The sensors were installed on the floor of the test bench (Figure 25 right). The MEMS
sensor plate was installed on the top of the IMV sensor to be as close as possible to
the same position. Two other tri-axial MEMS sensors were installed for comparison test
as well, which were designed for floor and backrest measurements.
Figure 25. Japan NIOSH 6-axis test bench (left) and installed sensors (right). The MEMS and
IMV sensors were installed to the same location to compare results.
Tests were made to validate the measurement equipment: 1) for determining the
practical sensitivity and noise characteristics of the system, 2) for determining the
sensor configuration to detect rotational vibrations (Table 9). Also a comparison of the
82
r.m.s. values was made to determine the practical error margin when analysing
comfort.
Tests consisted of stimuli having different amplitudes and frequencies separately for
each axis (fore-and-aft, lateral, vertical, roll, pitch and yaw) and for combination of one
translational and one rotational axis. Based on these tests the basic characteristics of
the sensor and measurement device (e.g. sensitivity and noise) was concluded and the
sensor configuration for calculating rotational directions was evaluated. Not all possible
combinations were tested, but at least two different amplitudes and frequencies for all
possible combinations of axes were used.
Table 9. Summary of the test stimuli conducted for evaluating the accuracy of the developed
sensor configuration. At least two frequencies and amplitudes were used for all possible
combinations of axes (all six axes separately and combination of one translational and rotational
axis).
Excitation signals Axes Amplitudes

1, 4, 8, 16, 40 and 80 Hz (sinusoidal), 0.25, 0.5, 1.0
Fore-and-aft, lateral, vertical, roll, pitch and yaw 2
0-20 Hz (random) and 2.0 m/s
The analysis included comparison of the sensor signals in time and frequency domain
and the differences of the frequency weighted r.m.s. values. The most important goal
was to evaluate the error margin of the r.m.s. values, because that in practice defines
the usability of the equipment.
Time domain analysis was not the most convenient way to compare signals, but
showed visually any discrepancies (e.g. Figure 26 left). A more efficient way was using
the frequency domain, in this case power spectral density (PSD) (e.g. Figure 26 right).
The results showed that both sensors corresponded well to the stimulus. Because IMV
had 1024 Hz sampling rate and the developed prototype equipment had 260 Hz
sampling rate, the signal from IMV was resampled to 260 Hz. The resampling process
was not exact so small differences can be seen in the time domain. However, the
resampled signals corresponded to each other very well and no significant differences
were found based on the time and frequency domain analyses. IMV had better noise
ratio, which was expected, because the prototype equipment did not yet include
efficient noise reduction, and because IMV has low noise piezoelectric sensors. The
noise of the developed equipment was due to a noisy power source (i.e. 50 Hz peak in
the spectrum) and because the sensor is ratiometric the output signal is influenced by
the noise of the input voltage. This issue was known and can be solved by introducing
a high precision voltage regulator.
83
Random vibration 0-20Hz 1
Random vibration 0-20 Hz vertical direction
10
2.5 MEMS MEMS
IMV IMV
0
10
2
-1
1.5 10
1 -2
10
0.5 -3
10
m /s 2
0
-4
10
-0.5
-5
10
-1
-6
10
-1.5
-7
10
-2
-8
10
3.42 3.44 3.46 3.48 3.5 3.52 3.54 3.56 3.58 0 10 20 30 40 50 60 70 80 90 100
Time (samples) 4 Frequency
x 10
Figure 26. Random vibration in time domain (left) and frequency domain (right) from the test
bench for MEMS and IMV.
Based on the frequency domain information, a transfer function estimate was

calculated to provide frequency response information between the two sensors (Figure
27). Between 0 and 20 Hz the signals from the sensors had highest differences in 10
Hz area (+8 % to -4 %). In other areas the difference was much smaller. In the 0 to 20
Hz area the mean ratio was 1.0091 and standard deviation 0.029.
Figure 27. A transfer function estimate comparing MEMS to IMV for vertical axis. Same analysis
was made for all axes for each stimulus.
The last procedure was to calculate frequency weighted r.m.s. values and to compare
the differences between them (Figure 28). For the example stimulus the MEMS sensor
84
showed 0.661 m/s2 and IMV 0.694 m/s2 for vertical direction. The difference was less
than 5 %. The same procedure was conducted for all analysed stimuli. A linear
trendline of the r.m.s. values shows that in general IMV gives 5.9 % higher values than
MEMS.
Comparison of RMS values
0,700
y = 1,0586x
0,600
0,500
IMV RMS (m/s^2)
0,400
0,300
0,200
0,100
0,000
0,000 0,100 0,200 0,300 0,400 0,500 0,600 0,700
MEMS RMS (m/s^2)
Figure 28. Comparison of the r.m.s. values of IMV and MEMS for all analysed stimuli. R.m.s.
values are calculated for combining three translational and three rotational axes based on the
overall vibration total value of the ISO 2631-1 standard.
The error margins found comparing to the commercial equipment were small (< 5 %)
and can be considered as sufficient for practical measurements, because other error
factors in field measurements are much higher. The technical specifications in ISO
8041 alone allow errors from ±10% to even up to ±27% (Mansfield 2003). Although the
commercial measurement equipment should give systematic results, in a study the
difference between measurement systems were concluded to be even 25% of each
other (Tyler and Darlington 2004). The accuracy was within the required limits even at
very small amplitudes (e.g. 0.06 m/s2).
4.3.2.4 Summary
A 6-axis seat pad was designed, constructed and validated using a 6-axis test bench at
JNIOSH, Kawasaki, Japan, and by comparing results to those obtained using a
commercially available system, compliant with ISO 8041. Small tri-axial MEMS-sensors
were used. A circuit board was designed including also anti-aliasing filters.
The developed prototype was the first test version and thus did not include all the
necessary components of commercial measurement equipment. Because the purpose
was mainly to show the usability of less expensive sensors for practical measurements,
a commercial development kit was used for data recording. The results showed that
85
using simple and less expensive components it is possible to realise a 12-axis
measurement system. The next step was to further develop the current equipment for
enabling practical field measurements.
4.3.3 Prototype II: Experiment for validating equipment for field

measurements
4.3.3.1 Purpose
The equipment of the first prototype was still impractical and the sensor signals needed
to be noise controlled. Thus a second prototype was needed. The goal for the second
prototype was to allow practical field measurements, so the equipment was designed to
be simple and small.
A digital version of the same MEMS sensor (prototype I) was chosen to minimise the
noise and need for additional components. The resolution of the digital version is
dependent on the chosen amplitude range and the performance of the noise reduction.
The internal A/D conversion resolution of the sensor is 12 bits, thus the maximum
theoretical resolution is 1/4096 of the amplitude range (in reality the sensor does not
use the whole A/D-range optimally, so the practical sensitivity is lower). There were
four possible amplitude ranges to choose from (Table 10). Using the maximum digital
resolution an optimal sensitivity was calculated. The ±1.5 g versions were chosen for
the seat pad and ±3.0 g for the floor and backrest, because it is rare to exceed 10 m/s2
acceleration on the seat, but floor vibrations can be much higher. The amplitude range
for the backrest sensor was chosen to be same as for the floor sensor for practical
purposes.
Table 10. The theoretical sensitivity (i.e. smallest detectable change in vibration amplitude) for
different amplitude ranges of the selected MEMS sensor.
2 2
Amplitude range (g) Amplitude range (m/s ) Sensitivity (m/s )
±1.5 29.43 0.0072
±2.0 39.24 0.0096
±3.0 58.86 0.0144
±6.0 117.72 0.0287
The sensors and the sensor configuration were designed based on the experience of
the first prototype. This time all sensors of the seat pad were integrated in to the same
86
circuit board, which also included all the other components of the device (Figure 29).
Only floor and backrest sensors had separate housings.
The sensors used in the second prototype were digital, thus no separate A/D
component was included as the sensors provided a digital signal directly. The digital
signals from the six 3-axis sensors were recorded using an FPGA chip, which sampled
at 400 Hz frequency. The raw acceleration data was saved to four 64 Mbit memory
chips. This meant that the equipment could save up to 24 minutes of raw acceleration
data from six 3-axis sensors at 400 Hz sampling rate without the need for external data
acquision.
The measurement procedure was designed to be very simple: while the device is
powered it is measuring and recording data. After the measurement the data is
downloaded from the device to PC using a terminal program and RS-232 port.
Figure 29. The 6-axis seat sensor and data acquisition system, which includes also memory and
processor for acquiring and calculating sensor data.
As the sensor configuration was tested previously, the calibration procedure of the
second prototype focused on the accuracy of the whole equipment. The experiments
were conducted in a laboratory in Finland. Three tests were conducted; 1) noise, 2)
sensitivity and 3) sensor calibration test. The noise test was conducted when the
sensors were resting on a table. The results showed that the noise was less than the
smallest resolution of the values (< 0.01 m/s2).
87
Sensitivity was tested and calibrated using the gravity effect. The sensors are
capacitive, so they detect the earth’s gravity. When the sensor is turned around its axis,
the signal value changes based on the gravity, thus the effect of gravitation can be
used to calculate the sensor’s practical sensitivity and to conduct static calibration
(Table 11). Because the information from the sensor was digital, the sensitivity in this
case means the same thing as resolution, and the values are in 12 bit integer.
Table 11. The sensitivity of prototype II for all sensors and axes based on the gravity test.
Sensor Axis Value (+g) Value (-g) Span (value) Span (gravity) Sensitivity (m/s2)
1 x 2044.5 1222.8 821.6 +0.5g...-0.5g 0.0119
1 y 2124.1 1298.2 825.9 +0.5g...-0.5g 0.0119
1 z 2963.6 1297.2 1666.4 +g...-g 0.0118
2 x 2056.6 1218.2 838.4 +0.5g...-0.5g 0.0117
2 y 2078.8 1287.8 791.0 +0.5g...-0.5g 0.0124
2 z 2921.3 1286.6 1634.7 +g...-g 0.0120
3 x 2027.9 1194.8 833.1 +0.5g...-0.5g 0.0118
3 y 2266.3 1464.9 801.4 +0.5g...-0.5g 0.0122
3 z 3102.9 1465.4 1637.4 +g...-g 0.0120
4 x 2052.5 1225.2 827.3 +0.5g...-0.5g 0.0119
4 y 2168.0 1357.5 810.5 +0.5g...-0.5g 0.0121
4 z 2996.2 1354.6 1641.6 +g...-g 0.0120
5 x 2033.0 1497.3 535.6 +0.5g...-0.5g 0.0183
5 y 2562.3 1487.5 1074.8 +g...-g 0.0183
5 z 2519.3 1433.7 1085.5 +g...-g 0.0181
6 x 2040.1 1473.3 566.8 +0.5g...-0.5g 0.0173
6 y 2602.8 1518.6 1084.3 +g...-g 0.0181
6 z 2581.2 1493.9 1087.3 +g...-g 0.0180
88
Figure 30 shows the gravitation test in time domain for one of the sensors for vertical
direction. The average value of the stabilised signal was calculated for both positions.
The smaller value was then subtracted from the larger one and the result was divided
by the gravity span.
Figure 30. Validation of sensitivity of prototype II by using gravitation test for a vertical axis of
one of the sensors.
4.3.3.4 Summary
The second prototype was developed for field measurements requiring minimal
components. The tests were conducted to determine the validity of the equipment for
whole-body vibration analysis. The results showed that the practical resolution of the
system including noise characteristics (less than 0.02 m/s2) was well enough for
analysing the discomfort in reasonable accuracy, and that the equipment can be used
for field measurements.
4.3.4 Commercial version
4.3.4.1 Purpose
It was previously concluded that there were no twelve axes measurement systems
commercially available in practice. One of the goals of the thesis work was to develop
measurement equipment supporting the full method, which would be commercially
viable. The purpose was to use the knowledge from the prototypes to produce a
89
commercial version of the measurement equipment. This version was developed with a
Finnish technology company Vibsolas, which is specialised in motion sensors and
applications (http://www.vibsolas.com).
The version is based on the same sensor type and configuration used in prototype II,
but it was decided to make a more universal sensor system, thus an interface to a
general data acquisition system was made. V12SP sensor system comprises from
three sensor boards:
• Seat pad sensor (V6SP) including four 3-axis accelerometers (Figure 31);
o Can detect translational and rotational acceleration ranges up to 6 g;
• Backrest sensor (V3SP) including one 3-axis sensor (Figure 32);
o Detects translational acceleration;
• Floor sensor (V3SP) including one 3 axis sensor (Figure 32);
o Detects translational acceleration.
The sensor configuration was designed for calculating translational and rotational
accelerations from the seat and translational accelerations from the backrest and floor.
The rotational accelerations are detected based on the sensor configuration selected
earlier.
Figure 31. The commercial 6-axis seat sensor V6SP board and mechanical housing.
90
Figure 32. The commercial 3-axis backrest and floor sensor V3SP (left: electronic board, right:
mechanical housing).
The analog outputs from the sensors are anti-alias filtered using 100 Hz single pole
filter. Recommended sampling rate is 10-20 times the cut-off frequency.
4.3.4.3 Improvements for the second prototype
V12SP was developed for whole-body vibration discomfort analyses and to comply with
ISO 2631-1 (1997) requirements. It was designed to use with any data acquisition
system (DAQ) that can process analog signals between 0-5 Volts.
Table 12 shows the sensitivities of the V6SP sensors (1.5 g and 3.0 g versions). The
sensitivities were calibrated using a gravity test with National Instruments NI-DAQ
system (power supply 12V, A/D input range 0-5V). The sensitivities might not be similar
in another data acquisition system.
91
Table 12. The sensitivities of the commercial V6SP sensor for 1.5 and 3.0 g versions using
National Instruments NI-DAQ system with 5 V input.
Sensor Axis Sensitivity 1.5 g (V/g) Sensitivity 3.0 g (V/g)
A1 x 1.34461 0.62034
A1 y 1.33249 0.66061
A1 z 1.32970 0.67333
A2 x 1.33230 0.64181
A2 y 1.32400 0.67069
A2 z 1.33387 0.65580
A3 x 1.32690 0.64603
A3 y 1.33668 0.66508
A3 z 1.34603 0.66251
A4 x 1.34174 0.62246
A4 y 1.35110 0.66326
A4 Z 1.34132 0.66335
4.3.4.5 Summary
A commercial version of the 12-axis sensor system was developed and validated. The
system can be used with any data acquisition system with A/D inputs from 0 to 5 V.
The equipment can be bought through the company website at a reasonable price.
4.4 Developing and validating analysis software

The process of measuring and analysing the 12-axis vibration for discomfort evaluation
is described in Figure 33. The whole process involves several steps from acquiring the
raw acceleration signal to the calculation of the final value. Some of the steps are done
in the measurement equipment, such as detecting the acceleration signals, anti-
aliasing the signals, A/D converting and recording them in the memory. After
downloading the data to a computer it needs to be unit converted, rotational directions
calculated, frequency weighted, r.m.s. and CSF (Crest Factor; r.m.s. peak value)
values calculated, multiplying factors applied, and finally the point vibration total value
and overall vibration total value calculated. The final version of the analysis program
was made using a Python development environment.
92
Acceleration signal from sensors
A/D conversion
Measurement
equipment
Direction of analysis
Normalisation
Recording to memory
Converting values to SI-units
Applying frequency weighting
Calculation rotational axes
Computer
Calculating r.m.s. and CSF values
Applying multiplying factors
Calculating point vibration total values
Calculating an overall vibration total value
Figure 33. The analysis procedure of the developed measurement equipment and software for
measuring, calculating and evaluating discomfort from whole-body vibration using the full
method of ISO 2631-1.
4.4.1 Converting to SI-units
Each sensor has a sensitivity range, which is noted in relation to the digital value and
acceleration (e.g. 864 counts/g for digital sensors and 560 mV/g for analog sensors).
The values from the sensor need to be converted to represent acceleration values
using this information. In the Table 11 there are calculated sensitivity values for each of
the sensor directions used in this study, based on the gravity test. The conversion is
93
done using equation 20, where the sensor value Ssensor is multiplied by the sensitivity
value:
RSI = S sensor ⋅ Sensitivity (20)
Figure 34 shows the change in the signal after unit conversion. At this point the data
has not been filtered yet.
2
Figure 34. Converting the raw acceleration data from A/D values to SI units (m/s ) (left: before
and right: after).
4.4.2 Frequency weighting
Each axis has a defined frequency weighting curve (i.e. filter). The standard gives the
values to create the filtering algorithms to represent the frequency weightings. There
are different methods to realise the filters. In this case the filters were developed in a
Python environment using digital IIR-filter design.
The developed filters can be validated in several ways. At first the evaluation can be
done by plotting developed and the standardized filters in the frequency domain.
Secondly weighting values at each 1/3 octave band can be compared and for third
r.m.s. and crest factor values can be calculated for the reference stimulus in the
standard (ISO 8041 2005). The developed filters are visually identical to the standard’s
filters (Figure 35 and Figure 36).
94
Figure 35. Frequency weighting curves Wk (left) and Wd (right) for both developed and
standard’s filters in frequency domain.
Figure 36. Frequency weighting curves Wc (left) and We (right) for both developed and
standard’s filters in frequency domain.
The developed filters comply with the tolerance requirements of ISO 8041 standard up
to 100 Hz (Table 13). The software uses sampled data up to 100 Hz and rest of the
data is cut-off by resampling process.
95
Table 13. Weighing values for the standard and developed weighting curves and errors (%)
1
between them at each /3 octave frequency.
Developed filter ISO 2631-1 Standard Error (%)
Freq (Hz) Wk Wd Wc We Wk Wd Wc We Wk Wd Wc We
0.100 31.5 63.1 63.0 63.2 31.2 62.4 62.4 62.5 1.0 1.1 1.0 1.1
0.125 50.3 100.7 100.6 101.0 48.6 97.3 97.2 97.5 3.5 3.5 3.5 3.6
0.160 80.2 160.6 160.3 161.3 79.0 158.0 158.0 159.0 1.5 1.6 1.5 1.5
0.200 121.5 243.7 243.2 245.3 121.0 243.0 243.0 245.0 0.4 0.3 0.1 0.1
0.250 181.8 365.3 364.0 368.5 182.0 365.0 364.0 368.0 -0.1 0.1 0.0 0.1
0.315 262.1 528.5 525.7 534.9 263.0 530.0 527.0 536.0 -0.3 -0.3 -0.2 -0.2
0.400 351.3 711.6 705.9 721.6 352.0 713.0 708.0 723.0 -0.2 -0.2 -0.3 -0.2
0.500 417.6 851.5 841.7 860.8 418.0 853.0 843.0 862.0 -0.1 -0.2 -0.2 -0.1
0.630 458.5 943.5 928.5 938.3 459.0 944.0 929.0 939.0 -0.1 -0.1 -0.1 -0.1
0.800 476.9 992.1 972.2 940.8 477.0 992.0 972.0 941.0 0.0 0.0 0.0 0.0
1.000 482.4 1010.9 990.9 879.6 482.0 1011.0 991.0 880.0 0.1 0.0 0.0 0.0
1.250 484.6 1007.5 1000.2 772.2 484.0 1008.0 1000.0 772.0 0.1 0.0 0.0 0.0
1.600 494.3 968.3 1006.6 631.8 494.0 968.0 1007.0 632.0 0.1 0.0 0.0 0.0
2.000 531.4 890.2 1011.7 511.5 531.0 890.0 1012.0 512.0 0.1 0.0 0.0 -0.1
2.500 630.6 775.9 1017.1 409.0 631.0 776.0 1017.0 409.0 -0.1 0.0 0.0 0.0
3.150 804.1 641.9 1022.5 323.0 804.0 642.0 1022.0 323.0 0.0 0.0 0.0 0.0
4.000 967.3 511.8 1023.8 253.0 967.0 512.0 1024.0 253.0 0.0 0.0 0.0 0.0
5.000 1038.9 408.9 1012.8 201.6 1039.0 409.0 1013.0 202.0 0.0 0.0 0.0 -0.2
6.300 1054.4 322.9 974.1 159.5 1054.0 323.0 974.0 160.0 0.0 0.0 0.0 -0.3
8.000 1036.2 252.8 890.4 125.3 1036.0 253.0 891.0 125.0 0.0 -0.1 -0.1 0.2
10.00 987.8 201.3 775.0 100.0 988.0 212.0 776.0 100.0 0.0 -5.0 -0.1 0.0
12.50 900.9 160.4 644.8 79.9 902.0 161.0 647.0 80.1 -0.1 -0.4 -0.3 -0.3
16.00 765.7 124.8 509.2 62.2 768.0 125.0 512.0 62.5 -0.3 -0.2 -0.5 -0.4
20.00 631.4 99.4 405.6 49.6 636.0 100.0 409.0 50.0 -0.7 -0.6 -0.8 -0.8
25.00 506.7 79.0 321.2 39.5 513.0 80.0 325.0 39.9 -1.2 -1.3 -1.2 -1.1
31.50 397.4 62.1 251.3 31.0 405.0 63.2 256.0 31.6 -1.9 -1.7 -1.8 -1.8
40.00 305.9 48.1 193.9 24.0 314.0 49.4 199.0 24.7 -2.6 -2.6 -2.6 -2.7
50.00 236.5 37.4 150.3 18.7 246.0 38.8 156.0 19.4 -3.9 -3.6 -3.7 -3.7
63.00 176.0 28.0 112.2 14.0 186.0 29.5 118.0 14.8 -5.4 -5.1 -4.9 -5.6
80.00 120.0 19.1 76.6 9.6 132.0 21.1 84.4 10.5 -9.1 -9.5 -9.2 -9.0
100.0 69.6 11.1 44.5 5.6 88.7 14.1 56.7 7.1 -21.5 -21.3 -21.5 -21.4
96
The values for reference signal from ISO 8041 show practically same results for the
developed software (Table 14). This is the official test based on the standard to
validate the frequency weightings. Results show less than 1 % error for each test.
Table 14. Calculated r.m.s. values for the standard and developed frequency weightings using
guided reference frequencies.
Reference Reference ISO 2631-1 Developed

Application Weighting
frequency amplitude filter filter
Wc 0.5145 0.5144
Whole-body Wd 0.1261 0.1261

2
15.915 Hz 1.0 m/s
vibration We 0.06287 0.06286
Wk 0.7718 0.7716
Figure 37 shows the acceleration signal before and after the frequency weighting.
Because the standard’s filtering emphasises only lower frequencies the characteristics
of the signal can change significantly in the time domain.
Figure 37. An example of the effect of frequency weighting wd for the acceleration data in time
domain (left: before and right: after).
4.4.3 Root mean square values and crest factors
The r.m.s. calculation can be validated using reference signals. In this case the
reference signal was 1 Hz and 1 m/s2 sine wave (Table 15). In this case the r.m.s.
value should be 0.707 and the CSF should be 1.4142. The results showed a very good
match to ideal values, thus the software functions can be concluded to work properly.
Table 15. Validating the r.m.s. and CSF calculations using a reference signal.
Reference Reference Theoretical Theoretical Calculated Calculated

freq. amplitude r.m.s. CSF r.m.s. CSF
2 2 2
1 Hz 1 m/s 0.707 m/s 1.414 0.707 m/s 1.414
97
4.4.4 Point and overall vibration total values
Validating the functions, which calculate point and overall vibration total values
(Equations 13 and 14) is simple. The results were tested using Microsoft Excel function
for vector sum and comparison to the developed Python function. Results from the
calculations can be verified and determined that programmed functions work (Table
16).
Table 16. Reference values to validate if programmed algorithms to calculate point and overall
vibration total values work. Both values are calculated using a vector sum principle (Equation
13).
Results
Reference values
Excel Python
for x, y, and z
1,2,3 3.741657 3.741657
4,5,6 8.774964 8.774964
7,8,9 13.92839 13.92839
10,11,12 19.10497 19.10497
4.4.5 Calculating rotational axes
Because the chosen solution for recording rotational accelerations is based on

translational accelerometers, the rotational directions need to be calculated based on
at least two sensors. The rotational directions can be calculated using Equations 17-19.
There is a reference publication that has already validated the method (Yamashita and
Maeda 2003). The data from experiment I was used to validate the software.
4.4.6 Software development
The analysis software was first developed and tested. It was further developed using
Python programming language, thus a commercial and license free software was
produced (PyHuman package).
4.4.7 Summary
The validation of the analysis process and software was found to comply with the
standard. The frequency weighting curves were nominally identical up to 100 Hz, which
is the effective range of the system and exceeds the 80 Hz upper limit described in ISO
2631-1. Also the r.m.s. and CSF equations gave practically identical results compared
98
to ISO 8041 requirements. Based on the test it was concluded that the analysis is
accurate and complies with the standards’ requirements.
4.5 Conclusions
12-axis measurement equipment and analysis software were developed and validated.
The tests showed compliance with the standards’ requirements. The standards do not
include guidance on how to build and test the 6-axis seat pad sensor, but another
reference was used to realise it (Yamashita and Maeda 2003). Based on the results it
was concluded that the equipment and the software are accurate enough to be used
for the full discomfort evaluation based on ISO 2631-1 standard.
99
5 Chapter 5 - Field measurements:
contribution of twelve axes according
to the ISO 2631-1 method
5.1 Introduction
Because of the lack of usable measurement equipment and the confusing guidance of
the standardised method, there have been very few studies that have used the
standard ISO 2631-1 method to evaluate discomfort from whole-body vibration and
even fewer that have measured and analysed all twelve axes in the field. Thus there
has not been enough data to evaluate how the twelve axes are present in the field and
which measurement locations are important.
Typically 3-axis sensors are used for measuring vibration in the field, thus in practice
there is more interest in evaluating how many locations need to be measured, instead
of the individual axes. As rotational axes require more complex measurement
configuration, it is of interest to conclude if rotational axes can be completely neglected.
However, because of lack of publications analysing the effects of the axes in the field,
the relative importance of individual axes are also needed to be analysed, as well as
the effects of frequency weighting and multiplying factors for emphasising the axes and
locations. Additionally the usability of the r.m.s. method (i.e. crest factor values) and
possibilities to compensate missing locations using multiplying factors need
clarification.
Based on the literature review in Chapter 2, only two adequate publications were found
to have 12 axis data (Griffin 1990, Maeda 2004). However, it was found that neither
publication gave enough information to conduct all the necessary analyses to conclude
the relative effects of the axes. Also the data sets in the publications were relatively
small, thus additional measurements were considered necessary. The literature review
in Chapter 2 found also another 12-axis measurement study (Parsons and Griffin
1983), but the method they used is not the same as in the standard.
The relative importance of the axes is affected by the frequency weighting curves, the
multiplying factors and the root-sum-of-squares (r.s.s.) method for combining the axes
to point and overall vibration total values (PVTV and OVTV). The frequency weighting
models the response of the body in the frequency domain and the multiplying factors
define the relative importance of the axes in the same location when calculating
100
PVTVs. The differences between locations are emphasised when PVTVs are combined
to OVTV as PVTVs are multiplied to the second power (i.e. highest PVTV is
emphasised). Thus, the standard procedure might affect the results so that certain
axes and locations will always be most dominant (i.e. there is no need to measure
some axes in any environment). This needs to be verified.
The standard’s recommendation to use 1.4 multiplying factors for horizontal axes when
using only translational axes from the seat has no reference. There are no papers that
have validated the factors. The purpose of the factors is not clear, but it is assumed
that they are used to compensate for the missing backrest axes (i.e. OVTV using the
seat translational axes or seat and backrest translational axes should give similar
values). It needs to be validated for different work environments. It is also noted that no
compensating factors are provided for floor translational or seat rotational axes. Similar
factors should also be produced for them if they contribute to the OVTV.
For most work machines and vehicles the seat has a relatively rigid frame, thus
vibration measured from the seat surface should correlate to backrest vibration. Also
floor vibration can have a correlation to seat vibration. Depending on the suspension
and cushion type, the correlation may vary significantly. If comparison of OVTVs from
different measurements is done, then the axes, which have not been used, should be
compensated so that all results would equal the value using all axes. It would be easier
to have a compensating factor for PVTVs instead of each axis. If PVTVs of locations
have high correlation, then they can be compensated by a single multiplying factor.
The standard method assumes that the weighting procedure will produce a value that
is linked to the discomfort from vibration exposure. However, the frequency weighted
r.m.s. method produces the averaged value of each axis for the whole frequency range
from 0.5 to 80 Hz. It might be that an axis has dominant amplitude at a specific
frequency range, but is not dominant if the whole range is averaged, such that the axis
or PVTV could be neglected in the overall assessment. The discomfort might be more
a sum of combination of axes, which are dominant at each frequency range, than
based only on the dominant axis or location. It has been noted in previous studies that
amplitudes and frequency characteristics change the discomfort depending on whether
low or high frequencies dominate. This suggests that vibration has different effects
depending on the frequency range.
The conclusions from the previous publications and the analysis of the standard
weighting procedure lead to the following hypotheses:
101
• All twelve axes are not needed to be measured (i.e. not all measurement
locations) (e.g. (Wikstroem, et al 1991));
o Most likely at least rotational axes can be neglected;
• There is a correlation between some axes, thus their effect can be

approximated based on the other measured axes (i.e. the effect of some
measurement locations can be approximated by using a multiplying factor) (e.g.
(ISO 2631-1 1997));
• Most dominant axes will be translational axes from the seat and backrest fore-
and-aft (i.e. most dominant measurement location will be the seat surface) (e.g.
(Maeda 2004));
• Different axes are dominant at low frequencies than high frequencies (i.e.
weighting attenuates the axes differently);
• R.m.s. method is the best method to calculate vibration magnitudes of the axes
and r.s.s. for combining the axes for evaluating discomfort (e.g. (Parsons and
Griffin 1983)).
5.2 Goals of the field measurements

The goal is to analyse how the nine translational and three rotational axes, and the
three measurement locations, which are set by the standard (ISO 2631-1 1997), have
importance in real work environments. Based on the analyses, laboratory and field
trials can be designed and conducted to find out the relative effects on discomfort.
The purpose of the analyses is to find answers to the following questions:
• What are the dominant and negligible axes and locations in practice?
• Are the dominant axes and locations similar in all machines or do they change
between machines and environments?
• Does dominance in axes change for different frequency ranges?
• What are the effects of the ISO 2631-1 frequency weighting curves and
multiplying factors for emphasising axes and locations?
• Can the r.m.s. method be used to evaluate the effects of vibration (i.e. is crest
factor threshold exceeded)?
102
• Can vibration from another location be approximated by using a multiplying
factor, thus simplifying the number of the measured locations (e.g. the standard
suggests that the effect of seat backrest axes can be replaced by 1.4
multiplying factors for seat horizontal axes)?;
• Are results from the previous studies and the new field measurements similar
enough to have more general conclusions from the results?
The new field measurements are conducted to extend the published data sets. The
purpose is to measure 12-axis whole-body vibration exposure from typical mobile work
machines (e.g. wheel loader) and transport equipment (e.g. bus) in normal working
conditions. All new measurements are planned to be conducted in real working
conditions with real machine operators. At least two different work phases of each
machine and at least two repetitions of minimum of 3 minutes each is measured. If the
measured task is shorter than 3 minutes, then it is repeated as long as the minimum
measurement period is fulfilled.
5.3 Methods
5.3.1 Results from previous publications
The results of the 12-axis data found in the previous studies (Griffin 1990, Maeda
2004) were included in the analysis with the new field measurements. The frequency
weighted r.m.s. values for each axis was used. Neither study reported unweighted
r.m.s. values or crest factors, thus some analyses were not possible from the previous
publications.
5.3.2 Measurement procedure
Although the measurement plan included the measured machine types, the machine
model, manufacturing year, etc, were not chosen beforehand. These were determined
by the availability of the machine to be measured. Some of the machines (a bus, train
and car) were chosen because the other publications had measured them also. This
was done for including overlapping data with previous measurements for direct
comparison and reference.
Before the installation of the equipment, the information of the machine and
environment was documented. Also the sensors were calibrated with a gravity test
before each installation. The sensors were installed on a seat, backrest and floor using
adhesive tape (Figure 38). The equipment with the same setup (i.e. sensor
103
configuration) was used for all measurements. All measurements were recorded as raw
acceleration data from six 3-axis sensors. Four 3-axis sensors were used to measure
rotational and translational axes from the seat based on Figure 23, one 3-axis sensor
recorded translational axes from the floor and one 3-axis sensor recorded translational
axes from the backrest. The measurement procedure and purpose was explained to
the machine operator. The measurement itself was automatic, so that the operator did
not need to do anything except use the machine normally. Each operator signed a
consent form.
Figure 38. The sensors were installed to the floor, seat surface and backrest based on the
guidance of the ISO 2631-1 standard. The equipment used in the new field measurements were
based on the second prototype version (Chapter 4).
During the measurements the additional information of the work phases,

surface/terrain, speed, etc, were documented. Because the measurements for each
machine were conducted in the period of a few hours, the environmental conditions
(e.g. weather and surface) could not be chosen and are not comparable to other
measurements. Based on the sensor information the 12-axis data was calculated in the
developed PC environment (Chapter 4). The measurements were done in real working
conditions comprising 1-7 repetitions of a single work phase for 3-10 minutes at one
time (Table 17).
104
Table 17. The measured machines and information from the new field measurements
conducted in this thesis. Each machine and work phase was measured in an authentic
environment, where the operator worked normally.
Duration per
Machine
Machine type Work phase Speed Repetitions repetition (min to
model
max)
Tractor #1 New Holland Moving gravel road 43 km/h 4 3-5 mins
Tractor #2 New Holland Moving asphalt road 43 km/h 4 3-5 mins
Tractor #3 New Holland Cultivating in field 12 km/h 2 3-5 mins
Excavator #1 Volvo Moving 5 km/h 2 3-5 mins
Excavator #1 Volvo Digging 0 km/h 5 3-5 mins
Train #1 Pendolino Moving 130 km/h 5 8-10 mins
Train #2 Pendolino Slowing down 130-0 km/h 1 10 mins
Train #3 Pendolino Accelerating 0-130 km/h 1 8 mins
Car #1 Volvo Moving asphalt road 60 km/h 3 5-10 mins
Car #2 Volvo Moving asphalt road 100 km/h 1 8 mins
Car #3 Volvo Moving stone road 30 km/h 2 10 mins
Harvester #1 Ponsse Moving in forest 3 km/h 6 5-10 mins
Harvester #2 Ponsse Harvesting trees 0 km/h 1 5 mins
Bus #1 Volvo Moving in city area 10-40 km/h 7 3-8 mins
Truck #1 Scania Moving in city area 10-50 km/h 6 3-8 mins
5.3.3 Signal processing
The raw acceleration data from the sensors was bandpass filtered from 0.5 to 80 Hz as
required in the standard, before the frequency weighting filters were applied. The
rotational axes were calculated based on the vibration data from the seat pad sensor
(see Figure 24 and Equations 17-19 in Chapter 4). The signal processing was
conducted using the previously validated software.
5.3.4 Vibration magnitudes
Vibration magnitudes were calculated for each axis with and without frequency
weightings using the r.m.s. method. Crest factors were determined for the weighted
values to analyse the applicability of the r.m.s. method. R.s.s. (Equation 14) were used
to combine axes from the same location to PVTVs and OVTVs with and without the
documented multiplying factors. Translational and rotational axes were calculated
separately from the seat surface as guided in the standard, thus creating four PVTVs:
105
1) seat translational axes (as), 2) backrest translational axes (ab), 3) floor translational
axes (af) and 4) seat rotational axes (ar).
The four most likely measurement scenarios were created for calculating OVTVs: 1)
only seat translational axes are measured (as), 2) seat and backrest translational axes
are measured (asb), 3) seat, backrest and floor translational axes are measured (asbf)
and 4) all twelve axes are measured (av). Scenario 1 represents a typical and simplest
measurement scenario, where only seat translational axes are measured using a
standard seat pad. Scenarios 2 and 3 are more difficult to realise, but still possible as
many measurement systems support multiple sensor inputs (normally up to 16 input
channels). Scenario 4 is the most difficult to realise, as it requires using also sensors
for measuring rotational axes. However, it is the recommended scenario in ISO 2631-1
if there is no information on the importance of the axes in the environment.
It is also important to note that in most studies scenario 1 has been used without
multiplying factors (i.e. 1.0) as described in the standard’s main text. However, the
standard notes that 1.4 multiplying factors should be used for seat horizontal axes, if
the backrest axes are not measured. Based on the results from many studies it is clear
that this note has not been considered and rarely have the multiplying factors been
used when evaluating discomfort (the same multiplying factors are mandatory when
evaluating the health effects). It is of interest to analyse to role of the compensating
multiplying factors, thus two version from scenario 1 was created: as1.0 (no multiplying
factors) and as1.4 (with 1.4 multiplying factors for horizontal axes). In other scenarios
PVTV is calculated without the multiplying factors (as1.0) as they all include the backrest
axes.
It is important to note that PVTV of seat translational axes will produce the same value
as OVTV in scenario 1. In this case the 1.4 multiplying factors should always be used.
Also the assumption is that if other locations are used then they include already
locations from the previous scenarios, thus no measurement is made only from the
backrest or floor.
5.3.5 Relative magnitude and contribution of the values
The relative magnitudes and contribution of all twelve axes and locations were
analysed. The relative magnitude ari (in percentage) can be calculated as:
ai
ari = ⋅ 100% (21)
amax
106
where ai is the r.m.s. value of a location (i.e. PVTV) or an axis i and amax is the highest
r.m.s. value of OVTV or the axes.
The relative contribution of a component to the OVTV aci (in percentage) can be
calculated as:
2
a 
aci =  i  ⋅ 100% (22)
 av 
where av is OVTV and ai is the r.m.s. value of a PVTV or an axis i.
5.3.6 Dominant axis at different frequency range
The dominant axis at each 1/3 octave frequency was determined for each machine
type for the new field measurements using three setups: 1) frequency weighting and
multiplying factors, 2) frequency weighting without multiplying factors and 3) without
frequency weighting and multiplying factors. The change in the dominant axis and the
influence of frequency weighting and multiplying factors were evaluated. The
evaluation was made by calculating the r.m.s. value for all twelve axes for each 1/3
octave frequency. The highest value was the dominant axis in that frequency.
5.3.7 Effects of frequency weighting and multiplying factors to the

relative contribution of axes and locations
The relative magnitude and contribution of each axis and PVTV to OVTV was
calculated with and without frequency weighting. Additionally the effect of multiplying
factors were analysed based on comparing the relative contribution of PVTVs with and
without the factors.
5.3.8 Effect of the 1.4 multiplying factors for seat horizontal axes to
compensate backrest axes
The multiplying factors for seat horizontal axes were validated by calculating the
factors, which would equalise as and asb:
asb2 − awz
2
asb = as ⇒ asb = k 2 ⋅ awx
2
+ k 2 ⋅ awy
2
+ awz
2
⇒k= (23)
2
awx + awy
2
where asb is OVTV of seat and backrest translational axes, as is PVTV of seat axes with
multiplying factors, awx, awy and awz are the frequency weighted r.m.s. values of seat
translational axes x, y and z and k is the multiplying factor for horizontal axes.
107
5.3.9 Calculating compensating factors for overall vibration total values
using a different number of axes
The analysis of correlation between number of included axes was conducted by

comparing OVTVs of different scenarios by calculating Spearman rho (r2). The purpose
of this method was not to analyse the true correlation in frequency domain (i.e.
coherence), but the practical correlation when using the standard method for selecting
the appropriate number of axes for calculating OVTV. Additionally compensating
multiplying factors were calculated for scenarios 1-3 that would equal av using the
following principle:
av = ki ⋅ ai (24)
Where ki is a compensating factor for scenario i, and ai is OVTV based on scenario i.
5.4 Results
5.4.1 Results from the previous publications
Two publications were found that reported data which was appropriate for the study. In
this case only results from these two publications were taken, because no other well
enough documented papers were found. Previous publications had partly similar
machines and work environments. Griffin (1990) showed data from three machines,
which were each measured for about 60 seconds. Maeda (2004) published results from
seven machines. Maeda’s measurements were from 10 to 120 minutes and 4-5
repetitions (Maeda 2007).
Some analyses were not possible from the results of the previous studies as they only
reported the frequency weighted r.m.s. values.
5.4.2 Comparison of vibration magnitudes on different machines
OVTV (av) varied significantly based on the machine type and environment (Table 18).
The smallest OVTVs occurred for the train and car. However, the car showed more
variability between the values, because of different surface types. The train had fairly
similar track conditions for all of the measurements; thus the values showed more
consistency. The highest OVTVs were for the forklift and harvester. The harvester
exhibited large variation in the values depending on whether it was moving (#1) or
harvesting (#2). Moving in the forest caused three times higher values. In all machines
108
the r.m.s. values from seat rotational and floor translational axes were relatively small
compared to OVTV.
The frequency weighted r.m.s. values for each axis showed the highest magnitudes for
seat vertical and backrest fore-and-aft axes (Tables 18 and 19). Vehicles driven on
smooth surfaces (e.g. car, bus and train) showed small magnitudes for horizontal axes
compared to the vertical axis. Vehicles driven on more uneven surfaces (e.g. work
machines) had more equal magnitudes for all translational axes. Also rotational
motions were significant for the aforementioned machines. The results show
consistency amongst different studies.
109
2
Table 18. The frequency weighted point vibration total values (m/s ) for seat translational axes
with 1.0 multiplying factors (as1.0), backrest translational axes (ab), floor translational axes (af)
and seat rotational axes (ar) with respective multiplying factors. The frequency weighted overall
2
vibration total values (m/s ) for seat translational axes with 1.4 multiplying factors (as1.4), seat
and backrest translational axes (asb), seat, backrest and floor translational axes (asbf) and all
axes (av) with respective multiplying factors for all measurements. Relative differences (%) are
compared to the overall vibration total value using all axes (av).
Point vibration total values Overall vibration total values Relative differences (%)
Measurer Machine as1.0 ab af ar as1.4 asb asbf av as1.0 as1.4 asb afsb
Maeda Bus 0.537 0.265 0.364 0.155 0.558 0.648 0.667 0.717 75 78 90 93
Maeda Taxi 0.280 0.115 0.205 0.094 0.300 0.347 0.359 0.377 74 79 92 95
Maeda Bulldozer 0.999 0.565 1.147 0.355 1.261 1.521 1.561 1.661 60 76 92 94
Maeda Excavator 0.574 0.205 0.602 0.196 0.750 0.832 0.855 0.879 65 85 95 97
Maeda Tractor 0.801 0.315 0.562 0.232 0.915 0.979 1.006 1.055 76 87 93 95
Maeda Combine 0.604 0.258 0.396 0.432 0.734 0.722 0.841 0.880 69 83 82 96
Maeda Monorail 0.139 0.059 0.100 0.046 0.156 0.172 0.178 0.187 74 84 92 95
Maeda Fork 1.378 0.399 0.934 0.449 1.454 1.665 1.724 1.770 78 82 94 97
Griffin Bus 0.450 0.262 0.362 0.201 0.522 0.578 0.612 0.665 68 79 87 92
Griffin Car 0.430 0.136 0.269 0.344 0.451 0.507 0.613 0.628 69 72 81 98
Griffin Train 0.318 0.016 0.244 0.078 0.393 0.401 0.409 0.409 78 96 98 100
Marjanen Car #1 0.235 0.078 0.175 0.124 0.267 0.293 0.318 0.327 72 82 89 97
Marjanen Car #2 0.245 0.080 0.183 0.128 0.273 0.306 0.332 0.341 72 80 90 97
Marjanen Car #3 0.619 0.267 0.426 0.332 0.684 0.751 0.821 0.863 72 79 87 95
Marjanen Train #1 0.142 0.045 0.094 0.052 0.157 0.170 0.178 0.184 77 86 93 97
Marjanen Train #2 0.117 0.038 0.074 0.039 0.127 0.139 0.144 0.149 79 85 93 97
Marjanen Tractor #1 0.839 0.155 0.828 0.393 1.064 1.179 1.242 1.252 67 85 94 99
Marjanen Excavator #1 0.638 0.162 0.642 0.377 0.772 0.905 0.981 0.994 64 78 91 99
Marjanen Excavator #2 0.533 0.136 0.496 0.194 0.645 0.728 0.753 0.765 70 84 95 98
Marjanen Harvester #1 1.242 0.182 1.123 0.211 1.626 1.675 1.688 1.698 73 96 99 99
Marjanen Harvester #2 0.448 0.147 0.337 0.101 0.587 0.561 0.570 0.588 76 100 95 97
Marjanen Bus #1 0.556 0.066 0.388 0.112 0.732 0.678 0.687 0.690 81 106 98 100
Marjanen Truck #1 0.411 0.065 0.317 0.163 0.509 0.519 0.544 0.548 75 93 95 99
110
2 2
Table 19. The frequency weighted r.m.s. values (m/s or rad/s ) without multiplying factors for
seat translational axes (Xs, Ys and Zs), seat rotational axes (Roll, Pitch and Yaw), backrest
translational axes (Xb, Yb and Zb) and floor translational axes (Xf, Yf and Zf) for all
measurements.
Measurer Machine Xs Ys Zs Roll Pitch Yaw Xb Yb Zb Xf Yf Zf
Maeda Bus 0.124 0.097 0.513 0.156 0.583 0.385 0.319 0.178 0.610 0.128 0.288 0.333
Maeda Taxi 0.079 0.075 0.258 0.131 0.193 0.080 0.219 0.110 0.228 0.124 0.116 0.210
Maeda Bulldozer 0.646 0.449 0.615 0.698 0.623 1.200 1.383 0.538 0.348 0.520 0.872 0.620
Maeda Excavator 0.458 0.179 0.297 0.192 0.385 0.275 0.734 0.242 0.138 0.500 0.392 0.288
Maeda Tractor 0.334 0.302 0.663 0.331 0.513 0.560 0.653 0.380 0.218 0.356 0.456 0.455
Maeda Combine 0.353 0.240 0.427 0.328 0.353 0.255 0.458 0.226 0.255 0.776 1.264 0.553
Maeda Monorail 0.036 0.063 0.119 0.033 0.128 0.100 0.093 0.084 0.133 0.056 0.124 0.078
Maeda Fork 0.247 0.405 1.294 0.320 0.625 1.170 1.091 0.534 0.493 0.356 0.464 1.063
Griffin Bus 0.161 0.218 0.359 0.334 0.375 0.075 0.379 0.216 0.418 0.192 0.156 0.478
Griffin Car 0.080 0.114 0.407 0.166 0.213 0.055 0.265 0.174 0.350 0.360 0.372 0.798
Griffin Train 0.082 0.221 0.214 0.025 0.008 0.005 0.246 0.236 0.208 0.080 0.112 0.175
Marjanen Car #1 0.096 0.087 0.196 0.108 0.096 0.039 0.192 0.098 0.170 0.118 0.134 0.289
Marjanen Car #2 0.088 0.085 0.213 0.100 0.120 0.037 0.207 0.092 0.159 0.096 0.144 0.301
Marjanen Car #3 0.172 0.242 0.543 0.338 0.401 0.119 0.470 0.271 0.370 0.303 0.403 0.767
Marjanen Train #1 0.026 0.064 0.124 0.057 0.065 0.031 0.105 0.060 0.076 0.079 0.078 0.109
Marjanen Train #2 0.023 0.044 0.106 0.048 0.057 0.027 0.082 0.041 0.070 0.050 0.058 0.085
Marjanen Tractor #1 0.413 0.526 0.506 0.195 0.235 0.057 0.884 0.745 0.538 0.479 0.838 0.775
Excavator
Marjanen 0.287 0.339 0.458 0.175 0.266 0.260 0.764 0.304 0.315 0.759 0.613 0.717
#1
Excavator
Marjanen 0.240 0.283 0.382 0.179 0.176 0.149 0.529 0.426 0.363 0.368 0.343 0.369
#2
Harvester
Marjanen 0.794 0.718 0.630 0.039 0.209 0.799 1.268 0.633 0.908 0.498 0.385 0.350
#1
Harvester
Marjanen 0.284 0.263 0.225 0.212 0.147 0.061 0.308 0.412 0.254 0.183 0.191 0.192
#2
Marjanen Bus #1 0.345 0.343 0.270 0.080 0.092 0.103 0.400 0.366 0.301 0.229 0.230 0.193
Marjanen Truck #1 0.247 0.183 0.273 0.080 0.093 0.076 0.332 0.271 0.273 0.322 0.268 0.314
5.4.3 Applicability of the r.m.s. method
If the crest factor value exceeds the threshold value of nine, the standard suggests that
the r.m.s. method underestimates the effects of vibration on discomfort (because of
high shocks). In almost all cases (11 out of 14) at least one axis showed crest factor
higher than the threshold value (Table 20). The floor axes most commonly exceed the
threshold. If only the seat and backrest axes were measured (Scenario 2) then 8 out of
111
14 showed crest factors over the threshold. If seat axes alone were used (Scenario 1)
then 5 out of 14 exceeded the threshold. The analysis could be done only from the new
field measurements as the crest factors were not available from the previous studies.
Table 20. The crest factor values for frequency weighted seat translational axes (Xs, Ys and
Zs), seat rotational axes (Roll, Pitch and Yaw), backrest translational axes (Xb, Yb and Zb) and
floor translational axes (Xf, Yf and Zf). Values exceeding the threshold limit 9 of ISO 2631-1 are
highlighted.

Marjanen Car #1 6.1 4.8 4.9 4.8 4.4 4.6 5.4 5.4 4.5 5.9 6.4 5.5
Marjanen Car #2 5.0 4.5 9.2 5.5 4.7 4.5 5.9 7.1 3.9 9.8 4.6 11.6
Marjanen Car #3 6.0 6.2 7.1 5.9 5.5 8.7 5.5 4.8 6.4 13.6 7.4 7.5
Marjanen Train #1 5.0 4.6 5.1 4.9 4.1 9.9 5.0 4.6 5.1 5.5 4.7 5.9
Marjanen Train #2 9.3 5.1 6.1 5.2 4.6 4.0 4.9 5.1 5.9 6.4 6.2 5.9
Marjanen Excavator #1 5.7 5.3 6.6 7.3 5.3 6.8 6.5 7.0 7.8 7.6 7.4 10.0
Marjanen Excavator #2 6.6 10.5 9.0 6.4 7.4 9.9 12.0 8.6 9.0 16.9 14.8 16.9
Marjanen Harvester #1 6.8 5.4 9.5 7.6 6.0 6.5 7.7 6.0 8.3 22.4 22.5 22.5
Marjanen Harvester #2 6.0 5.1 5.7 5.5 6.2 4.1 5.6 4.7 11.5 7.8 7.2 5.9
Marjanen Bus #1 6.0 6.9 12.2 5.6 6.4 5.6 8.0 5.3 5.9 9.5 10.7 12.5
Marjanen Truck #1 8.2 6.1 8.9 6.9 7.3 6.5 9.6 6.6 6.5 8.3 10.7 9.9
Average 6.2 5.7 7.3 5.9 6.1 6.3 6.9 5.7 6.4 9.4 8.7 9.5
5.4.4 Relative magnitude and contribution of point vibration total values
For most of the measured machines PVTV from the seat translational axes were the
highest (Table 21). PVTV of the backrest axes were dominant in few cases (3 out of
25), but the seat rotational and floor axes had no dominant PVTVs. In general they
were less than 50 % of the dominant PVTV. In 7 out of 25 cases the seat rotational
axes could have been neglected based on the standard recommendation (less than 25
% of the dominant PVTV can be neglected). For floor the translational axes 3 out of 25
PVTVs were below 25 % margin.
112
Table 21. The relative magnitudes (%) of point vibration total values for seat translational axes
(as1.0), backrest translational axes (ab), floor translational axes (af) and seat rotational axes (ar)
with respective multiplying factors. The location which was the highest is treated as the most
important point vibration total value (i.e. 100 %). The rest of the axes are compared to the most
important point vibration total value (embolden). Underlined values are below 25 % of the
highest value, thus could be neglected based on the standard guidance.
Measurer Machine as1.0 ab ar af
Maeda Bus 100 68 49 29
Maeda Taxi 100 73 41 34
Maeda Bulldozer 87 100 49 31
Maeda Excavator 95 100 34 33
Maeda Tractor 100 70 39 29
Maeda Combine 100 66 43 71
Maeda Monorail 100 72 42 33
Maeda Fork 100 68 29 33
Griffin Bus 100 81 58 45
Griffin Car 100 62 32 80
Griffin Train 100 77 5 24
Marjanen Car #1 100 74 33 53
Marjanen Car #2 100 75 32 52
Marjanen Car #3 100 69 43 54
Marjanen Train #1 100 67 32 36
Marjanen Tractor #1 100 99 19 47
Marjanen Excavator #1 99 100 25 59
Marjanen Harvester #1 100 90 15 17
Marjanen Bus #1 100 70 12 20
Marjanen Truck #1 100 77 16 40
Average 99 79 30 39
113
The relative contribution of PVTVs to OVTV shows that the seat PVTV had the largest
contribution on average (Table 22). It constituted on average about half of the
contribution. From the rest of PVTVs only backrest axes have any significance.
2
Table 22. The relative contribution (%) of frequency weighted point vibration total values (m/s )
for seat translational axes (as1.0), backrest translational axes (ab), floor translational axes (af)
and seat rotational axes (ar) with respective multiplying factors to the overall vibration total value
of all axes.
Measurer Machine as1.0 ab ar af
Maeda Bus 56 26 14 5
Maeda Taxi 55 29 9 6
Maeda Bulldozer 36 48 12 5
Maeda Excavator 43 47 5 5
Maeda Tractor 58 28 9 5
Maeda Combine 47 20 9 24
Maeda Monorail 55 29 10 6
Maeda Fork 61 28 5 6
Griffin Bus 46 30 15 9
Griffin Car 47 18 5 30
Griffin Train 61 36 0 4
Marjanen Car #1 51 28 6 14
Marjanen Bus #1 65 32 1 3
Marjanen Truck #1 56 34 1 9
Average 52 33 6 9
114
The effect of PVTV can be more clearly visualised from Figure 39. Seat and backrest
translational PVTVs contribute 4/5th of OVTV. The results are consistent for all
measurements.
Figure 39. Relative contribution (%) of frequency weighted point vibration total values with
respective multiplying factors to the overall vibration total value.
5.4.5 Relative magnitude and contribution of individual axes
The most dominant axis was the seat vertical for most of the environments (Table 23).
The second most dominant was backrest fore-and-aft and for third both seat fore-and-
aft and lateral axes had practically the same importance. The least important axis was
seat yaw axis. The seat vertical was the most dominant for even surfaces and
machines that are used in transportation, such as buses and cars. The few exceptions
were tractor and combine (measured by Maeda). The backrest fore-and-aft was
generally dominant in machines that were used on rough surfaces, such as the
bulldozer, excavators, and harvesters.
115
Table 23. The relative magnitudes (%) of the frequency weighted seat translational axes (Xs, Ys
and Zs), seat rotational axes (Roll, Pitch and Yaw), backrest translational axes (Xb, Yb and Zb)
and floor translational axes (Xf, Yf and Zf) with multiplying factors. The axis which was the
highest is treated as the most important axis (i.e. 100 %). The rest of the axes are compared to
the most important axis (embolden).

Maeda Bus 24 19 100 19 45 15 50 17 48 6 14 26
Maeda Taxi 31 29 100 33 30 6 68 21 35 12 11 33
Maeda Bulldozer 58 41 56 40 23 22 100 24 13 12 20 22
Maeda Excavator 78 30 51 21 26 9 100 21 9 21 17 20
Maeda Tractor 50 46 100 32 31 17 79 29 13 13 17 27
Maeda Combine 83 56 100 49 33 12 86 26 24 45 74 52
Maeda Monorail 30 53 100 18 43 17 62 35 45 12 26 26
Maeda Fork 19 31 100 16 19 18 67 21 15 7 9 33
Griffin Bus 45 61 100 60 42 4 84 30 47 13 11 53
Griffin Car 20 28 100 26 21 3 52 21 34 22 23 78
Griffin Train 37 100 97 7 1 0 89 53 38 9 13 32
Marjanen Car #1 49 45 100 35 20 4 78 25 35 15 17 59
Marjanen Car #2 41 40 100 30 23 4 78 22 30 11 17 57
Marjanen Car #3 32 45 100 39 30 4 69 25 27 14 19 57
Marjanen Train #1 21 52 100 29 21 5 68 24 24 16 16 35
Marjanen Train #2 21 41 100 29 21 5 62 19 26 12 14 32
Marjanen Tractor #1 58 74 72 17 13 2 100 53 30 17 30 44
Marjanen Excavator #1 47 55 75 18 17 9 100 25 21 31 25 47
Marjanen Excavator #2 57 67 90 27 17 7 100 50 34 22 20 35
Marjanen Harvester #1 78 71 62 2 8 16 100 31 36 12 9 14
Marjanen Harvester #2 100 93 79 47 21 4 87 72 36 16 17 27
Marjanen Bus #1 100 99 78 15 11 6 93 53 35 17 17 22
Marjanen Truck #1 91 67 100 19 14 6 97 50 40 30 25 46
Average 54 55 88 25 22 9 83 33 31 18 20 37
116
The vertical axis from the seat and fore-and-aft axis from the backrest showed
significant contributions to OVTV on average (Table 24 and Figure 40). Horizontal axes
from the seat showed also meaningful contribution, but the rest of the axes showed
contribution of 6 % or less. Compared to relative magnitudes (Table 23) the dominant
axis showed the same trend where the seat vertical is dominant for even surfaces and
lighter machines, and the backrest fore-and-aft is dominant for rougher surfaces.
Excavators and tractors showed highest dominance in the backrest fore-and-aft axis. In
other machines the seat vertical was most likely the dominant axis. The seat rotational,
backrest lateral and vertical, and floor translational axes had little or no significant
contribution to OVTV.
117
Table 24. The relative contribution (%) of frequency weighted seat translational axes (Xs, Ys
and floor translational axes (Xf, Yf and Zf) to the overall vibration total value with respective
multiplying factors.
Most important
axis
Maeda Bus 3 2 51 2 11 1 13 2 12 0 1 3 Zs
Maeda Taxi 4 4 47 5 4 0 21 2 6 1 1 5 Zs
Maeda Bulldozer 15 7 14 7 2 2 44 3 1 1 2 2 Xb
Maeda Excavator 27 4 11 2 3 0 45 2 0 2 1 2 Xb
Maeda Tractor 10 8 40 4 4 1 25 3 1 1 1 3 Zs
Maeda Combine 16 7 24 6 3 0 17 2 1 5 13 6 Zs
Maeda Monorail 4 11 40 1 7 1 16 5 8 1 3 3 Zs
Maeda Fork 2 5 53 1 2 2 24 2 1 0 0 6 Zs
Griffin Bus 6 11 29 10 5 0 21 3 6 1 0 8 Zs
Griffin Car 2 3 42 3 2 0 11 2 5 2 2 26 Zs
Griffin Train 4 29 27 0 0 0 23 8 4 0 0 3 Zs
Marjanen Car #1 9 7 35 4 1 0 22 2 4 1 1 12 Zs
Marjanen Car #2 7 6 39 3 2 0 24 2 3 0 1 12 Zs
Marjanen Car #3 4 8 39 6 3 0 19 3 3 1 1 13 Zs
Marjanen Train #1 2 12 45 4 2 0 21 3 3 1 1 6 Zs
Marjanen Train #2 2 9 51 4 2 0 19 2 4 1 1 5 Zs
Marjanen Tractor #1 10 19 16 1 1 0 30 9 3 1 3 6 Xb
Marjanen Excavator #1 9 11 21 2 1 0 36 3 2 4 2 9 Xb
Marjanen Excavator #2 10 12 25 3 1 0 29 10 4 1 1 4 Xb
Marjanen Harvester #1 22 18 14 0 0 1 36 3 5 1 0 1 Xb
Marjanen Harvester #2 25 18 15 5 1 0 16 12 4 1 1 2 Xs
Marjanen Bus #1 22 26 16 1 1 0 20 6 5 1 1 2 Ys
Marjanen Truck #1 20 13 24 2 1 0 22 5 5 2 2 5 Zs
Average 11 11 30 3 2 0 25 4 4 1 2 6 Zs
118
Figure 40. The relative contribution (%) of frequency weighted seat translational axes (Xs, Ys
and floor translational axes (Xf, Yf and Zf) to the overall vibration total value with respective
5.4.6 Effect of frequency weighting to the contribution of axes
The results show that the weighted OVTV was on average 46 % of the unweighted
value (Table 25). The influence of the frequency weighting was most significant for
vehicles operating on smooth surfaces (e.g. car and train). Tractors and harvesters
exhibited vibrations that were mostly present in the frequency range where attenuation
is smallest, thus the unweighted and weighted values had the smallest difference.
For individual axes, the largest attenuation occurred for rotational axes (Table 25). The
roll, pitch and yaw axes were less than 20 % of their unweighted values, while other
axes were 50 to 70 %. The most unaffected axes to frequency weighting were backrest
fore-and-aft (88 %) and seat vertical (71 %). Unweighted axes showed that roll and
pitch axes are the most dominant without weighting.
119
Table 25. The relative difference ratios (%) of r.m.s. values with and without frequency
weighting for seat translational axes (Xs, Ys and Zs), seat rotational axes (Roll, Pitch and Yaw),
backrest translational axes (Xb, Yb and Zb) and floor translational axes (Xf, Yf and Zf), and for
overall vibration total values for seat translational axes with 1.0 (as1.0), seat and backrest axes
(asb), all translational axes (afsb), and to all twelve axes (av) The percentage is ratio of the
weighted value to the unweighted value.
Relative difference (%)
Measurer Machine Xs Ys Zs Roll Pitch Yaw Xb Yb Zb Xf Yf Zf as1.0 asb afsb av
Marjanen Car #1 38 26 69 12 8 6 87 36 51 55 47 69 41 51 56 27
Marjanen Car #2 39 27 75 11 10 6 89 36 49 54 50 74 45 55 59 29
Marjanen Car #3 32 32 76 15 17 10 92 42 46 66 64 76 46 55 62 34
Marjanen Train #1 12 30 80 12 8 7 82 23 34 65 36 47 35 40 42 23
Marjanen Train #2 11 33 77 13 8 7 76 25 35 49 34 50 35 42 42 20
Marjanen Tractor #1 54 62 74 13 26 13 91 61 54 73 81 87 61 67 72 61
Marjanen Excavator #1 39 37 78 16 8 12 83 33 36 71 59 72 44 53 60 34
Marjanen Excavator #2 58 64 76 26 20 19 88 62 56 77 76 76 65 68 71 56
Marjanen Harvester #1 86 86 55 35 19 44 93 85 78 67 66 57 78 81 79 67
Marjanen Harvester #2 78 74 74 34 28 20 94 77 69 62 61 66 76 77 74 61
Marjanen Bus #1 77 76 62 17 17 17 93 83 82 65 65 65 74 79 76 60
Marjanen Truck #1 57 45 59 14 10 12 85 66 63 69 69 69 54 63 65 47
Average 52 53 71 18 15 16 88 54 57 66 62 66 57 63 65 46
Power spectral densities from weighted and unweighted acceleration data are
presented in Figure 41 and Figure 42. Both figures are from a car driven on a
cobblestone road (two separate data sets). Figures show spectra of the two
measurements (red) and an averaged spectrum (blue). The figures show that Wk and
Wc frequency weighting curves have less attenuation on the frequency band, thus seat
vertical, backrest fore-and-aft and floor translational axes are closer to unweighted
spectra than other axes. Especially We show attenuation, which significantly changes
the spectra of rotational axes. Because of the weighting the axes show most of the
energy in low frequency area (i.e. below 10 Hz). For unweighted data the energy is
spread in wider frequency range, especially for rotational axes.
120
Figure 41. Frequency spectra (dBm/Hz) of twelve unweighted axes for a car driven on a
cobblestone road (frequency range from 0 to 80 Hz). Blue line shows the mean spectra of two
separate measurements (red lines).
Figure 42. Frequency spectra (dBm/Hz) of twelve weighted axes for a car driven on a
cobblestone road (frequency range from 0 to 80 Hz). Blue line shows the mean spectra of two
separate measurements (red lines).
121
5.4.7 Effect of multiplying factors to the contribution of axes
PVTV from the seat (as) had smaller weighted r.m.s. value without multiplying factors,
because of the 1.4 multiplication factors for horizontal axes (Table 26). For other
combinations of axes OVTV was smaller with the multiplying factors (this is evident as
all the factors for additional axes are less than one). The combined value (asb) of the
seat and backrest was 81 % smaller. The seat, backrest and floor axes combined (afsb)
were 67 % smaller and OVTV (av) 64 % smaller. It was clear that the multiplying factors
attenuated rotational and floor axes significantly. This effect was systematic for all
environments.
122
2
Table 26. Overall vibration total values of frequency weighted r.m.s. values (m/s ) with and
without multiplying factors, and relative differences of values with factors to without factors.
With factors Without factors Relative differences
Measurer Machine as1.4 asb afsb av as1.0 asb afsb av as asb afsb av
Maeda Bus 0.558 0.648 0.667 0.717 0.537 0.891 1.002 1.231 104 % 73 % 67 % 58 %
Maeda Taxi 0.300 0.347 0.359 0.377 0.280 0.436 0.513 0.569 107 % 80 % 70 % 66 %
Maeda Bulldozer 1.261 1.521 1.561 1.661 0.999 1.822 2.176 2.655 126 % 83 % 72 % 63 %
Maeda Excavator 0.750 0.832 0.855 0.879 0.574 0.973 1.197 1.301 131 % 86 % 71 % 68 %
Maeda Tractor 0.915 0.979 1.006 1.055 0.801 1.122 1.342 1.577 114 % 87 % 75 % 67 %
Maeda Combine 0.734 0.722 0.841 0.880 0.604 0.831 1.787 1.869 122 % 87 % 47 % 47 %
Maeda Monorail 0.156 0.172 0.178 0.187 0.139 0.229 0.278 0.323 112 % 75 % 64 % 58 %
Maeda Fork 1.454 1.665 1.724 1.770 1.378 1.902 2.256 2.636 106 % 88 % 76 % 67 %
Griffin Bus 0.522 0.578 0.612 0.665 0.450 0.753 0.925 1.055 116 % 77 % 66 % 63 %
Griffin Car 0.451 0.507 0.613 0.628 0.430 0.639 1.145 1.178 105 % 79 % 54 % 53 %
Griffin Train 0.393 0.401 0.409 0.409 0.318 0.511 0.557 0.558 124 % 79 % 73 % 73 %
Marjanen Car #1 0.267 0.293 0.318 0.327 0.235 0.361 0.496 0.518 114 % 81 % 64 % 63 %
Marjanen Car #2 0.273 0.306 0.332 0.341 0.245 0.370 0.507 0.532 111 % 83 % 65 % 64 %
Marjanen Car #3 0.684 0.751 0.821 0.863 0.619 0.902 1.287 1.395 111 % 83 % 64 % 62 %
Marjanen Train #1 0.157 0.170 0.178 0.184 0.142 0.201 0.254 0.270 111 % 85 % 70 % 68 %
Marjanen Train #2 0.127 0.139 0.144 0.149 0.117 0.164 0.200 0.216 108 % 84 % 72 % 69 %
Marjanen Tractor #1 1.064 1.179 1.242 1.252 0.839 1.526 1.965 1.990 127 % 77 % 63 % 63 %
Marjanen Excavator #1 0.772 0.905 0.981 0.994 0.638 1.087 1.628 1.679 121 % 83 % 60 % 59 %
Marjanen Excavator #2 0.645 0.728 0.753 0.765 0.533 0.937 1.125 1.163 121 % 78 % 67 % 66 %
Marjanen Harvester #1 1.626 1.675 1.688 1.698 1.242 2.092 2.212 2.362 131 % 80 % 76 % 72 %
Marjanen Harvester #2 0.587 0.561 0.570 0.588 0.448 0.728 0.798 0.841 131 % 77 % 71 % 70 %
Marjanen Bus #1 0.732 0.678 0.687 0.690 0.556 0.833 0.915 0.929 132 % 81 % 75 % 74 %
Marjanen Truck #1 0.509 0.519 0.544 0.548 0.411 0.653 0.837 0.849 124 % 79 % 65 % 65 %
Average 0.672 0.729 0.763 0.785 0.672 0.900 1.137 1.232 119 % 81 % 67 % 64 %
123
A multiplying factor was calculated for the seat translational axes, which would produce
the same OVTV compared to including the backrest axes. In ISO 2631-1 this value is
defined as 1.4, but based on these results, the multiplying factor varied for different
measurements from of 1.3 to 2.5 (Table 27). On average it was 1.693. For the combine
(Maeda), the harvester (Marjanen) and the bus (Marjanen) the standard multiplying
factor overestimated the effect and for the rest it underestimated it.
2
Table 27. Comparison of overall vibration total values (m/s ) of seat translational axes with 1.4
multiplying factors (as1.4) and with seat and backrest translational axes (asb). and calculation of
factor which would equal as and asb. Difference (%) is as1.4 to asb.
Measurer Machine as1.4 asb Difference (%) Factor
Maeda Bus 0.558 0.648 86 2.52

Maeda Taxi 0.300 0.347 86 2.13
Maeda Bulldozer 1.261 1.521 83 1.77
Maeda Excavator 0.750 0.832 90 1.58
Maeda Tractor 0.915 0.979 93 1.60
Maeda Combine 0.734 0.722 102 1.36
Maeda Monorail 0.156 0.172 91 1.71
Maeda Fork 1.454 1.665 87 2.21
Griffin Bus 0.522 0.578 90 1.67
Griffin Car 0.451 0.507 89 2.17
Griffin Train 0.393 0.401 98 1.44
Marjanen Car #1 0.267 0.293 91 1.68
Marjanen Car #2 0.273 0.306 89 1.80
Marjanen Car #3 0.684 0.751 91 1.75
Marjanen Train #1 0.157 0.170 92 1.69
Marjanen Train #2 0.127 0.139 92 1.80
Marjanen Tractor #1 1.064 1.179 90 1.59
Marjanen Excavator #1 0.772 0.905 85 1.76
Marjanen Excavator #2 0.645 0.728 89 1.67
Marjanen Harvester #1 1.626 1.675 97 1.45
Marjanen Harvester #2 0.587 0.561 105 1.33
Marjanen Bus #1 0.732 0.678 108 1.28
Marjanen Truck #1 0.509 0.519 98 1.44
Average 0.672 0.729 93 1.69
124
5.4.8 The dominant axis at different 1/3 octave frequencies
The results show that for a car the dominant axis in low frequencies, for weighted and
multiplied data, varied between backrest fore-and-aft and seat vertical axis, while floor
axes dominated over 10 Hz frequencies (Figure 43). If the same data was calculated
without multiplying factors, then pitch axis was dominant in the low frequency range
from 1-2 Hz and floor axes started to dominate already at 7 Hz. The unweighted data
shows that rotational axes, especially roll, were the most dominant practically
throughout the whole frequency range.
Figure 43. Dominant axes at each 1/3 octave frequency from a car driven on a cobblestone
road. The top plot (red) represents frequency weighted acceleration data with multiplying
factors, middle plot (blue) represents weighted data without multiplying factors and bottom plot
(blue) unweighted data without multiplying factors.
125
If evaluating machines driven in fairly smooth environments (bus, car and train), they all
show differing patterns for the dominant axis when using frequency weighting and
multiplying factors (Figure 44). The only strong similarity was found with the
unweighted data, where the rotational axes dominated practically in all frequencies.
Seat and backrest axes dominated for all machines.
Figure 44. Dominant axes from bus (upper left), car (upper right) and train (below). The top plot
(red) represents frequency weighted acceleration data with multiplying factors, middle plot (blue)
represents weighted data without multiplying factors and bottom plot (blue) unweighted data
without factors.
On the other hand, machines that were driven in rougher surfaces (excavator,
harvester and tractor) had much stronger similarities than car, train and bus using the
standard’s weighting process (Figure 45). The backrest fore-and-aft axis dominated in
most frequencies below 10 Hz, and the seat vertical dominated frequencies higher than
10 Hz. Again the rotational axes dominated for unweighted data.
126
Figure 45. Dominant axes from excavator (upper left), harvester (upper right) and tractor
(below). The top plot (red) represents frequency weighted acceleration data with multiplying
factors, middle plot (blue) represents weighted data without multiplying factors and bottom plot
(blue) unweighted data without factors.
5.4.9 Compensating factors for point vibration total values
Correlation between scenario 1 (as) and other scenarios were calculated. The
correlation compared to scenario 2 (asb) was practically similar if 1.4 multiplying factors
were used for seat horizontal axes (Figure 46: left) or were not used (Figure 46: right).
Correlation was also calculated for comparing the effects of rotational and floor
translational axes to the correlation of OVTVs (Figure 47). The results show that
neither the floor axes (Figure 47: left) or the rotational axes (Figure 47: right) had a
practical effect to correlation. The results are for all studies and no differences were
found between the studies.
127
2
Figure 46. Correlation (Spearman R ) between frequency weighted overall vibration total value
of seat translational axes (as) (with 1.4 multiplying factors to seat horizontal axes is shown on
left and without the multiplying factors on right) and seat and backrest translational axes (asb)
with respective multiplying factors. The values are calculated based on the results from Maeda
(2004), Griffin (1990) and the new field measurements.
2
Figure 47. Correlation (Spearman R ) between frequency weighted overall vibration total value
of seat translational axes (as) with 1.4 multiplying factors to seat horizontal axes and seat,
backrest and floor translational axes with respective multiplying factors (afsb) (left), and
correlation for as and overall vibration total value using all twelve axes with respective
multiplying factors (av) (right). The values are calculated based on the results from Maeda
(2004), Griffin (1990) and the new field measurements.
As high correlation was found between scenario 1 and other scenarios, a

compensating factor for the scenarios could be calculated in order to force the gradient
of the regression line to 1 (Table 28). The factor is higher if less PVTVs are included. In
this case it was assumed that the order of including axes will be the same as the order
of the scenarios. For example if OVTV is calculated based on the seat translational
axes, the calculated OVTV (including 1.4 multiplying factors for seat horizontal axes) is
multiplied by 1.2 to compensate the effect of the other PVTVs.
128
Table 28. Compensating factors for different combinations of point vibration total values to
compensate the effects of other point vibration total values. The factor is used to multiply the
calculated overall vibration total value.
Compensation factor
as1.0 as1.4 asb asbf av
1.4 1.2 1.1 1.03 1.0
5.5 Discussion
The purpose of the analyses was to determine the magnitude of the twelve axes of
vibration present in field measurements. It is important to understand how the axes
contribute in real work environment and how the standard methods emphasise the
axes. Based on the standard guidance and previous literature there was not enough
data for conducting the analyses, thus new measurements were made.
Despite the method for measuring 12-axis vibration being available since 1987, only
two publications were found that had published 12-axis measurement results. Maeda’s
paper (Maeda 2004) was published in conference proceedings and Griffin’s results
were in his book “Handbook of Human Vibration” (Griffin 1990). This indicates that the
full method is not used widely.
For 12-axis measurements, guidance provided in ISO 2631-1 is incomplete. There is

substantial scope for interpretation which will affect OVTV result. If more axes are
measured, the vibration total value will increase. Thus if not measuring all axes, then
compensation factors should be created to estimate the overall exposure for the
environment. The data presented here could be used for such a purpose, where the
percentage contribution of typical exposures can be used to estimate the relative
contribution from that axis.
The discomfort evaluation method is based on using the r.s.s. method for combining
the values, which means that dominating axes are emphasized to the power of two. For
this reason the standard gives a possibility to neglect values less than 25 % of the
dominant value, because in reality their contribution will be small. For conducting
measurements in the field it is feasible to optimise the number of accelerometers and
channels used, thus it would be beneficial to have guidance on what axes are
important (i.e. needed to be measured) in reality. This guidance is not available in the
ISO 2631-1 (1997).
129
5.5.1 Relative contribution of locations and axes
The relative magnitudes and contribution analyses showed that the most important
locations are the seat and backrest (translational axes). For individual axes the seat
vertical and backrest fore-and-aft were the most important. The importance of the
backrest fore-and-aft axis has been documented before, but it has not been noted to be
more important than the other seat axes. In almost all cases the backrest fore-and-aft
was at least the second most important axis. Normally the seat vertical was dominant
for smoother surfaces and with machines that are used in relatively high speeds (e.g.
car and train). The backrest fore-and-aft axis was dominant for rough surfaces and
heavy work machines. In general both axes were among the three most dominant
axes. The effect of seat rotational and floor translational axes was small.
The relative contribution of the twelve axes confirmed that the few dominant axes will
determine OVTV. This was evident also for PVTVs. In all situations at least seat and
backrest translational axes need to be included for calculating OVTV. It is however
important to notice that OVTV can still be 10 to 20 % smaller when excluding the
rotational and floor translational axes and in most cases all PVTVs were more than 25
% of the highest PVTV, thus they were not suggested to be neglected. In some cases
this might lead to misleading conclusions of the discomfort level, especially if the value
is close to two different categories, or if comparing data sets which have been
produced using different combinations of accelerometers. This can be compensated for
by using a factor depending on the number of PVTVs included.
5.5.2 Applicability of the r.m.s. method
Based on the interpretation of the ISO 2631-1 it was assumed that the r.m.s. method
underestimates the effects of vibration even if only one axis exceeds the crest factor
limit. In 11 out of 14 cases the crest factor limit was exceeded for at least one axis. The
highest thresholds were for floor axes, which was expected, because in most cases the
floor has the least damping (i.e. occurrence of high peaks are more common). In
almost all cases only one or two axes exceeded the limit and in most cases it was the
additional axis, thus if only seat translational axes were measured then the threshold
would not have been exceeded for most measurements (5 out of 14). The crest factors
showed that it is almost inevitable that at least one of the twelve axes will exceed the
standard’s recommended limit value. The measurements showed that the crest factor
easily exceeds the threshold value of 9. If it is exceeded the standard recommends
calculating MTVV and/or VDV values, but there is no guidance on how to use them for
130
comfort evaluation. Also the standard does not explicitly say if all axes should be under
the limit for the r.m.s. method to work, or if just one axis exceeds the limit, even if this
axis might hardly contribute to the OVTV. Certainly the limit value of BS 6841 standard
(6) will be exdeeded in practically all cases.
These results indicate that the crest factor is not an effective method for evaluation of
vibration. The factor is based only on one single peak from the whole measurement
period, thus for similar vibration characteristics it is more likely to have a larger crest
factor if a longer period is measured (i.e. likeliness of one high peak increases). Also it
is more likely to have a high crest factor if a higher sampling rate is used for acquiring
the acceleration signal. In this case higher peaks can be detected, but the r.m.s. value
will be practically unaffected by the sampling rate. The frequency weighting somewhat
compensates this, but still other factors than the true content of the vibration will have a
significant effect on the factor. This has been also acknowledged by the international
community as the new draft amendment for the ISO 2631-1 standard notes the same
issue (ISO/TC 8/SC 4 2008).
5.5.3 Effects of frequency weighting
Frequency weightings significantly changed the relative emphasis of the frequencies

for each axis. The impact was significant especially for rotational axes. The weighted
rotational r.m.s. values were on average less than 20% of the unweighted values. The
frequency weighting curve for rotational axes (We) emphasises only 0.5 to 1 Hz narrow
frequency range. Frequencies over 1 Hz are significantly attenuated. For example at 10
Hz the weighting curve reduces the signal energy by 90%. The frequency spectra of
the unweighted acceleration of rotational axes showed that on average the frequency
range with vibration was higher than 1 Hz. The backrest fore-and-aft axis is weighted
by Wc curve. The curve emphasises frequency up to 8 Hz and is much less progressive
than We. Also Wd and Wk curves, which are used for the seat and floor translational
axes and backrest horizontal axes have higher frequency range than We. Thus it was
clear that the effect of the rotational axes was attenuated more than the other axes.
Frequency weighting affected the backrest fore-aft axis the least where the weighted
r.m.s. values were almost 90% of the unweighted value. The other axes were
attenuated by approximately 40 to 50 %.
A more detailed analysis showed that within the frequency range of the standard (0.5 to
80 Hz), multiple axes were dominant at different frequencies. In most cases the seat or
backrest translational axes were dominant below and the floor axes above 10 Hz.
131
However, the rotational axes were dominant for practically all environments in the
absence of the frequency weighting and multiplying factors. This again shows how
much the weighting influences the results.
Even though different machines exhibited different vibration values and relations, it was
clear that the frequency weighting process almost always attenuated the axes so that
either one of the seat translational or backrest fore-aft axis became dominant. Thus the
effect of terrain and machine type or model was marginal. The rotational axes were not
more dominant with machines that were driven on rougher surfaces or were larger (e.g.
harvester). It can be assumed that the rotational motion in these machines is shown
primarily as translational, as the point of measurement is at some distance above the
centre of rotation.
5.5.4 Effect of multiplying factors
The multiplying factors change the relative effects of the frequency weighted r.m.s.
values. For axes other than the seat translational the multiplication factors are less than
1. The backrest fore-and-aft factor is 0.8, but for the lateral and vertical axes the factors
are 0.5 and 0.4 respectively. This means that in many cases the fore-and-aft axis will
be the most dominant, even though it does not have the largest vibration magnitude.
The same is true also for the rotational and floor translational axes, where roll and
vertical axes have higher multiplying factors than other axes. Because of r.s.s. method
for combining the axes, the differences in the factors have a second power effect, thus
making the highest value even more significant. The axis with largest factor will most
likely dominate the whole PVTV and the other axes will have insignificant meaning. The
results showed that the axes with the highest multiplying factor were also the most
dominant axes after weighting and scaling. The factors also further minimised the
effects of rotational and floor axes.
The 1.4 multiplying factor used when only seat translational axes are measured is
based on the assumption that the backrest axes will have an effect on the discomfort.
No publications were found, which have validated the chosen factors and there is no
reference in the standard where the factors have been derived. In fact BS 6841 does
not have the 1.4 factors for replacing the backrest axes. It is not clear if the factor is
supposed to be used for simulating the effect of backrest vibration to discomfort
compared to without the backrest, or for simulating OVTV.
For the latter the results in the Table 27 showed that the multiplying factor that would
produce similar results for as and asb should be about 1.7 and that it varies even for the
132
same machine type for different surface or work phase. Thus there needs to be a study
to evaluate and validate the multiplying factor based on the field measurements. There
might be a different factor for certain types of work phases or machines. It is important
to notice, that the calculation of the factor based on the backrest axes is related to the
location of the sensor. Thus the installation will have an effect to the size of the factor.
It would also be helpful for the purpose of the multiplying factor to be specified in the
standard, such that comparisons between different measurement scenarios can be
made.
5.5.5 Need for measuring additional axes
For conducting measurements in the field it is beneficial to optimise the number of

measurement locations, thus it would be beneficial to have guidance on what axes are
important contributors to the OVTV and those which can be ignored. For 12-axis
measurements, guidance provided in ISO 2631-1 is incomplete. There is a substantial
scope for interpretation which will affect OVTV result. If more axes are measured,
OVTV will increase. Thus if not measuring all axes, then compensation factors should
be applied to estimate the overall exposure for the environment. The data presented
here could be used for such a purpose, where the percentage of contribution of typical
exposures can be used to estimate the contribution from that axis.
Assuming that the standardised method of vibration assessment is correct, then the
relative importance of the rotational axes from the seat and the translational axes from
the floor has little or no significance to the subjective feeling based on the results in this
paper, even though they are most likely larger than 25 % of the highest value. This
eases the measurement process significantly.
5.5.6 Approximating point vibration total values using multiplying factors
The standard does not provide guidance on compensating OVTV depending on how
many axes and locations are used. However, this has a direct effect on the size of the
value and to the assessment, as the subjective tables are based on absolute values.
Maeda (2004) noted that using just the seat translational axes will most likely produce
a value that is interpreted as more comfortable than using all axes. This was also
shown in this study. The seat translational axes alone produced OVTVs in average of
70 % of the value using all axes.
The different scenarios showed high correlation, thus a single value to compensate the
missing PVTVs could be used to estimate OVTV of all axes. This will improve the
133
comparison of values which are based on different number of axes and improve
assessment of discomfort, which is based on the absolute values like in the ISO 2631-1
standard.
5.5.7 Differences and similarities of new and previous measurements
Relative magnitudes from the bus, car and train showed good agreement with the seat
vertical being the most dominant and the backrest fore-and-aft being the second most
dominant for all studies. Overall, good agreement was found for the seat, backrest and
floor horizontal, and roll and yaw axes. Biggest disagreement was for the pitch axis,
where Maeda’s results showed fourth largest r.m.s. value and Griffin and Marjanen
showed ninth largest. On the other hand Maeda showed smaller relative magnitude for
the floor vertical than Griffin and Marjanen. For the relative contribution all sets showed
similar trend than with the relative magnitude. The seat vertical and backrest fore-and-
aft were clearly most dominant for the both sets. Also the same disagreements with the
pitch and backrest and floor vertical axes were found with the bus, car and train. For
the excavator and tractor the comparison between Maeda’s and Marjanen’s data show
good agreement with the backrest fore-and-aft and seat translational axes. The biggest
difference was for the roll, pitch, and backrest vertical axes. Also for the excavator and
tractor, similar agreements and differences were found with the relative contributions.
Compared to an earlier study by Parsons and Griffin (1983), the results were similar for
the ranking of most axes. The study used different cars and road types and concluded
that seat vertical and floor vertical axes had highest contribution. The seat and backrest
fore-and-aft axes were the next important, having thus similar findings than later
studies. However, the study was conducted without multiplying factors and using
weightings which were different from the standard.
5.5.8 Interpretation of the standard method for field measurements
The results showed that the three translational axes from the seat alone
underestimated the total vibration value in almost all cases, if compared to using all
twelve axes, even though the 1.4 multiplying factors were used. This will most likely
underestimate one or two categories of the subjective feeling than using the 12-axis
data. This has been indicated previously also (Maeda 2004). However, if the backrest
axes were included the difference was on average less than 10 %. This was because
the backrest fore-and-aft axis was either the most dominant or the second most
dominant in all measurements. The rotational axes from the seat and translational axes
134
from the floor had only marginal effect. In both cases the effect was only few
percentages. So it is suggested to measure at least the seat and backrest translational
axes.
5.5.9 Scope and limitations of the results
Even though the results presented in this paper are only from specific machines, which
mean that the levels of r.m.s. values cannot be generalized, the relative importance of
different axes can be considered to be representative of work machines in normal
situations. However, there is still only a small body of data reporting 12-axis
measurements. Some analysis could be made only from the new field measurements,
thus limiting the number of measured environments.
5.5.10 Suggestion for further work
Assuming that the standardized method of vibration assessment is correct, then the
relative importance of rotational axes from seat and translational axes from floor has
little or no significance to the subjective feeling based on the results in this chapter.
Thus it is most likely that they do not need to be measured. This eases the
measurement process significantly. However, there is still only a small body of data
reporting 12-axis measurements and so it is recommended that more 12-axis data is
gathered and reported so that the relative importance of the axes in further practical
environments can be determined.
The most recent study by Wyllie and Griffin (2007) suggests that the rotational axes
have more effect on discomfort than the standard method implies. The standard also
neglects the effect of phase between axes and this might emphasise the floor axes.
Thus a laboratory study to confirm the effects of different axes in a multi-axis
environment is needed.
5.6 Conclusions
The standardised method to evaluate discomfort from whole-body vibration
emphasises the dominant axes. It was found that:
• The most dominant axes in field were from the seat (vertical) and backrest
(fore-and-aft);
o Thus dominant PVTVs were seat and backrest translational locations;
135
• Because of the standard’s method, the effect of rotational and floor axes on the
vibration total value were negligible;
o And because of the weighting of frequency and axes, it is most likely that
in any type of environment the rotational and floor axes have marginal
effect, at least for the machine types measured here.
• The dominant axis changed for different frequency ranges, where the seat and
backrest axes were dominant at low frequencies (< 10 Hz) and floor axes at
high frequencies (> 10 Hz);
o The effect of floor axes were higher without multiplying factors, and
rotational axes were dominant without the frequency weighting;
• The crest factor limits the usability of the r.m.s. method, especially if all twelve
axes are used, because it is likely that at least one axis will show high crest
factor;
o If less axes are used (e.g. seat translational axes), then the crest factor
will be exceeded less often;
• Correlation analyses showed that a compensating factor can be produced to

estimate the effect of PVTV of another location, thus all results could be directly
compared even though different number of axes are included;
o Higher multiplying factors than 1.4 should be used to compensate

backrest axes, but the results showed large variation between
environments, thus a factor for PVTV is suggested instead of individual
axes;
• All studies showed similar results and trends, thus a more general notion can
be concluded about the role of the axes and standard weighting process to the
results.
The results suggested that there is no reason to measure the seat rotational and floor
translational axes. However, there is limited information on how important the locations
are for predicting discomfort of subjects. The relative contribution of axes to discomfort
need to be validated in a laboratory using test subjects.
136
6 Chapter 6 - Laboratory trial part I:
description and validation of trial
procedure, methods and results
6.1 Introduction
The results from the field measurements in Chapter 5 suggested that if the standard
method is correct, then the discomfort evaluation process can be simplified. Rotational
axes from the seat and translational axes from the floor provided negligible contribution
to overall vibration total value (OVTV) in all cases, because of the standard weighting
process. However, there is very little information on how the axes contribute to the
discomfort in practice. The studies conducted in a laboratory have normally used
stationary sinusoidal or flat band random vibration stimuli and rarely more than one
axis. The few studies, which have been conducted in the field, have used test tracks
with short duration stimuli and emphasis on shocks (Fairley 1995). Only one study
(Parsons and Griffin 1983) was found, which had measured all twelve axes and
subjects’ judgments, but the data was analysed using different weightings than in the
standard and without using multiplying factors. There is a need for evaluating how well
the full ISO 2631-1 method predicts discomfort in practice. This means non-stationary
and field-like stimuli for long durations.
Based on the selected methods, stimuli, trial procedure and subjects, the results from a
trial will have limitations. It is important to understand the limitations in relation to the
goals of the trial. Before any conclusive conclusions can be made and new methods
suggested, it is necessary to validate the results. The method selected for acquiring
and evaluating judgment data (i.e. continuous judgment method) has no references in
relation to vibration exposure, thus an analysis of the applicability of the method is
needed before further conclusions can be made. The stimuli, which is based on field
measurements, have uncontrolled frequency characteristics and non-stationary time
signal. Thus, the results might be limited to similar environments. Depending on the
selected trial procedure and guidance given to the subjects, and the characteristics of
the subjects, the same trial can produce different results. It has been a norm to use
students or laboratory staff in the trials, as they are easily available. However, there is
a suspicion that people who have no experience or who do not have to consider the
137
other aspects, such as costs of the trip, are not the best to evaluate discomfort for
general public (Oborne 1976).
Even though there is a clear need for more information on the applicability of the
standard method for evaluating discomfort, the results from the few related previous
studies does provide indications for using the standard method:
• More axes will improve correlation to discomfort (Griffin, et al 1982, Parsons

and Griffin 1983);
o However, seat translational axes will have the highest effect (Fairley
1995);
• Using the r.m.q. method provides better correlation than the r.m.s. method for
stimuli including shocks (Boileau, et al 1989);
• Higher multiplying factors for seat horizontal axes should be applied (i.e.
frequency weightings for horizontal axes will underestimate the subjective
discomfort judgment) (Maeda and Mansfield 2006b).
6.2 Goals of the trial

Based on the literature review in Chapter 2 and results from the field measurements a
laboratory trial was planned. The trial should be conducted in a multi-axis environment
using stimuli, which simulate vibration exposures experienced in the field, as there has
not been this kind of study before. The results should seek to validate the conclusions
from the previous studies and provide new information on the role of axes, frequency
weighting, multiplying factors and averaging method for improving the correlation to
discomfort in the field. The stimuli should include shocks, but also characteristics,
which are more common in road transportation (i.e. non-stationary vibration).
The goals and analyses of the laboratory trial are divided into three separate phases:
1. Validate the chosen methods for acquiring and analysing discomfort judgments
and the effects of stimuli and gender to the results (Chapter 6);
2. Validate the standard method (Chapter 7);
3. Improve the standard method (Chapter 8).
The first goal is to validate and verify that the chosen trial procedure, stimuli and
judgment method provides results that can be used to validate and improve the
standard. It is important to understand if the chosen stimuli order or subjects will affect
138
the results. Based on the requirements of the latter goals the following questions were
set for the validation of the trial methods (part I):
• Does the order of stimuli affect judgments?
• Do the characteristics of subjects affect judgments (i.e. weight, height, gender)?
• What is the best approach for averaging judgment value for best correlation?
6.3 Methods
6.3.1 Test environment
6.3.1.1 Multi-axis shaker
A 6 degrees-of-freedom shaker at Loughborough University was used to expose

subjects to multi-axis vibration stimuli, which simulate vibration from field environments
(Figure 48). The shaker can be excited using movement sampled at 50 Hz frequency.
Each axis has a separate column of displacement values, thus six tab delimited
columns comprise the data file. If acceleration or velocity data is to be used, a signal
conversion (i.e. integration) and resampling is required to transform the data to
displacement information.
Figure 48. Six-axis shaker at Loughborough University for conducting the multi-axis laboratory
trial.
139
6.3.1.2 Sensor setup and data acquisition
Accelerometers were installed based on the standard guidance. Figure 49 shows the
sensor layout for the shaker. A backrest sensor was installed half way up the backrest
(270 mm from SIP). A floor sensor was installed below the seat at the same vertical
axis as the seat sensor. The floor sensor was used to analyse the platform movement,
but was not used for analysing the judgments. It was reasonable to assume that feet
did not make any significant contribution to the judgment, and this was confirmed in
pilot work (Bhalchandra 2008). Also it was previously concluded from the field
measurements (Chapter 5) that the floor axes showed marginal contribution to the
OVTV. The subject’s feet rested on a footrest, which was adjusted based on the
subject’s height and did not move during the experiment.
3-axis sensor
6-axis sensor
150 mm 330 mm
270 mm
300 mm
510 mm 270 mm
1000 mm
3-axis sensor
Figure 49. Shaker dimensions and sensor layout for the laboratory trial (side view).
The data was measured using a 12-axis sensor system produced by Vibsolas Ltd (the
commercial version V12SP in Chapter 4) and National Instruments recording device.
The recording and visualisation program was realised using a Labview 7.1
development environment. For each trial, the value of discomfort judgment and the
twelve axes of the standard were recorded at sampling frequency of 1000 Hz. Six axes
from the seat, including three translational and three rotational axes, three translational
axes from the backrest and three translational axes from the floor were measured.
140
6.3.1.3 Stimuli
Stimuli were based on the acceleration data from the field measurements conducted in
Chapter 5. The purpose was to create an environment that simulated frequencies and
relative magnitudes of the axes that are present in the field, thus using an approach
with high ecological validity. The stimuli were created taking the limitations of the
shaker into account. This included band-pass filtering between 1 and 20 Hz. Each main
stimulus was chosen so that as many as possible different frequency contents that are
present in the field were covered (Table 29). A signal processing equalisation was
conducted to the original measured 12-axis data, so that the stimuli represented similar
characteristics to the original source.
Table 29. Main stimuli used in the trial to represent vibration from the field. Each stimulus was
based on a selected data sample of the field measurements in Chapter 5. Each main stimulus
lasted 15 seconds.
Main stimulus Simulated machine Work phase Terrain Speed
Cobblestone road –
A Car Moving 30 km/h
city
B Truck Moving Asphalt – city 30-60 km/h
C Forestry harvester Moving Forest 3 km/h
D Train Moving Rail track 140 km/h
E Excavator Digging Gravel road 0 km/h
As signal noise and errors cause problems in the integration process, highpass filtering
was applied for the converted acceleration data. In this case cut-off frequency for the
highpass filter was 1 Hz using a 12th pole filter (Butterworth). A 20th pole low pass filter
(Butterworth) at 20 Hz was also used, thus the controlled frequency range was
between 1 and 20 Hz. However, due to friction and valve properties, additional
frequencies above 20 Hz did exist (Figure 50). These frequencies could not be directly
controlled, but they were considerably smaller in amplitude than the stimuli frequency
characteristics. Frequencies below 10 Hz dominated the spectrum for each axis. For
fore-and-aft axis the difference was over 15 dB, for lateral axis over 5 dB and for
vertical axis over 10 dB.
141
Figure 50. Power spectra from the multi-axis test bench of the floor translational axes for the
train stimulus (left: fore-and-aft, middle: lateral and right: vertical axis).
Figures 50 - 54 show the frequency content of each main stimulus. The power spectra
of the main stimuli represent the actual measured acceleration from the seat of the test
bench. The spectra show that for translational axes the low frequencies (< 10 Hz)
dominate. For rotational axes higher frequencies were noted as well, which is partly
due to an interaction between the seat cushion and the sensor pad. All stimuli showed
non-flat spectra and frequency characteristics, which are typically seen in the field.
Figure 51. Unweighted power spectra of the main stimulus A (car) for each axis measured from
the seat surface of the test bench.
Figure 52. Unweighted power spectra of the main stimulus B (truck) for each axis measured
from the seat surface of the test bench.
142
Figure 53. Unweighted power spectra of the main stimulus C (harvester) for each axis
measured from the seat surface of the test bench.
Figure 54. Unweighted power spectra of the main stimulus D (train) for each axis measured
from the seat surface of the test bench.
Figure 55. Unweighted power spectra of the main stimulus E (excavator) for each axis
measured from the seat surface of the test bench.
143
For each main stimulus six variations were created that had different relative vibration
magnitudes for the axes in order to test the ISO 2631-1 method. The variations can be
used to analyse how discomfort is affected by the change of dominant axes, thus
testing the effect of the multiplying factors. As each of the main stimuli had different
frequency content, it was also possible to evaluate how similar OVTVs were perceived,
if the frequency content was different, thus testing the effect of frequency weighting.
In total 30 different stimuli were created (Table 30). At least one variation of each main
stimulus had little or no rotational vibration, but had similar OVTV than stimulus which
had rotational vibration. All stimuli lasted 15 seconds.
144
Table 30. Stimuli and control factors for the test bench for all stimuli variations.
Control factors
Main
Variation Note x y z Roll Pitch Yaw
stimulus
A 1 Car – original 1 1 1 1 1 1
A 2 Car - enhanced vertical 0.5 0.5 1 0.5 0.5 0.5
A 3 Car - w/o rotation 0.5 0.5 1 0 0 0
A 4 Car - enhanced rotation 1.5 1.5 1 1.5 1.5 1
A 5 Car - enhanced pitch 1.5 0.75 0.5 0.75 1.5 0.5
A 6 Car - enhanced fore-and-aft 1.5 0.75 0.5 0 0 0
B 1 Truck – original 0.9 1 0.6 1 0.8 1
B 2 Truck - enhanced lateral 0.45 1 0.3 0.5 0.5 0.5
B 3 Truck - enhanced lateral w/o rotation 0.45 1 0.3 0 0 0
B 4 Truck - enhanced vertical 0.9 0.5 1.2 1 0.8 1
B 5 Truck - enhanced vertical 0.45 0.25 1.2 0.5 0.4 0.5
B 6 Truck: w/o rotation 0.45 0.25 1.2 0 0 0
C 1 Harvester – original 1 1 1 1 1 1
C 2 Harvester - enhanced vertical 0.5 0.5 1 0.5 0.5 0.5
C 3 Harvester - enhanced vertical 0.5 0.5 1 0 0 0
C 4 Harvester - reduced all 0.5 0.5 0.5 0.5 0.5 0.5
C 5 Harvester - enhanced vertical 0.25 0.25 0.5 0.25 0.25 0.25
C 6 Harvester - w/o rotation 0.25 0.25 0.5 0 0 0
D 1 Train – original 0.5 0.5 0.5 1 1 1
D 2 Train - w/o rotation 0.5 0.5 0.5 0 0 0
D 3 Train - enhanced vertical w/o rotation 0.25 0.25 0.5 0 0 0
D 4 Train - enhanced pitch 0.6 0.25 0.25 0.5 1 1
D 5 Train - enhanced fore-and-aft 0.75 0.25 0.25 0 0 0
D 6 Train - enhanced fore-and-aft 0.75 0.125 0.125 0 0 0
E 1 Excavator – original 1 1 1 1 1 1
E 2 Excavator - enhanced vertical 0.5 0.5 1 0.5 0.5 0.5
E 3 Excavator - enhanced vertical 0.2 0.2 1 0.2 0.2 0.2
E 4 Excavator - enhanced fore-and-aft 1 0.2 0.2 0.2 0.2 0.2
E 5 Excavator - enhanced pitch 1 0.1 0.1 0.2 1.5 0.2
E 6 Excavator - enhanced rotation 1 0.1 0.1 1.5 0.2 0.2
145
6.3.2 Judgment method
Subjects evaluated discomfort of each stimulus using the continuous judgment method
(Kuwano and Namba 1985). The trial was planned to be conducted using cross-modal
matching of subjective judgment and visual line. Subjects were presented a discomfort
line (Figure 56 left) that they could control in real-time using a rotary control (Figure 56
right). There was no numerical scale indicating the length of the line. The test subjects
were asked to adjust the line so that it corresponded to discomfort judgment between
“no discomfort” and “high discomfort”. The test subjects were guided to evaluate the
instantaneous discomfort for each stimulus separately. At the end of each stimulus the
test subjects reset the judgment by turning the indicator to “no discomfort” position.
Figure 56. The visual information presented to participants to judge discomfort based on the
continuous judgment of line length (left) and a rotary control for evaluating discomfort (right).
The continuous judgment method has not been commonly used in vibration research,
but has been used in noise research. In the previous studies a delay between the
stimulus and the response has been noted (see Chapter 2). There was also a
suspicion that some seconds from the beginning of each stimulus should be neglected
as the judgment started from “no discomfort” position, thus a visual inspection of the
judgment data was considered necessary. An arithmetic mean of the judgment was
calculated similarly as in the noise studies to produce a single judgment value for each
stimulus.
6.3.3 Subjects
The experiment used 22 subjects (12 males and 10 females) (Table 31). The average
age of the subjects was 22 years (three subjects were above 30 years). The average
height was 181 cm for men and 164 cm for women. Average weight was 74.8 kg and
59.6 kg respectively. Each subject gave an informed consent to participate in the trial.
The trial was approved by Loughborough University Ethical Advisory Committee.
146
Table 31. Information on the test subjects used in the trial. The stimuli were randomised and
three sequences (a, b and c) were created. The order of sequence for each subject was
changed.
Subject Sequence order Age (y) Gender Height (cm) Weight (kg)
1 a,b,c 23 Male 178 60
2 b,c,a 22 Male 180 72
3 c,a,b 30 Male 175 85
4 a,c,b 36 Male 192 80
5 c,b,a 18 Female 160 57
6 b,a,c 24 Male 174 60
7 a,b,c 20 Male 186 85
8 b,c,a 31 Male 171 66
9 c,a,b 18 Male 180 84
10 a,c,b 18 Female 174 56
11 c,b,a 18 Female 163 65
12 b,a,c 20 Female 167 60
13 a,b,c 19 Female 155 46
14 b,c,a 18 Female 173 63
15 c,a,b 18 Male 182 71
16 a,c,b 18 Male 180 75
17 c,b,a 18 Male 194 79
18 b,a,c 18 Male 182 81
19 a,b,c 29 Female 153 47
20 b,c,a 18 Female 164 67
21 c,a,b 19 Female 165 65
22 a,c,b 22 Female 167 70
Average 22 173 68
6.3.4 Study procedure
Before the trial, test subjects completed an instruction sheet, health screen
questionnaire and consent form. The test subjects were instructed to sit on the seat in
a comfortable upright posture and leaning against the backrest (Figure 57). A test
sequence of five stimuli from the lowest vibration magnitude to the highest was used to
familiarise subjects with the vibration before the trials began and to allow for training of
controlling the judgment line on the screen. Each test subject was then exposed to
three randomised sequences of all 30 stimuli (90 stimuli in total). Between sequences
there was a break where the test subjects were asked to dismount from the seat and
147
move around in the laboratory for five minutes, in order to minimise the effects of
fatigue.
As each sequence took 10 minutes and there was a 5 minute break between the
sequences, the trial procedure took 40 minutes for each subject. The subjects used a
seat belt for safety purposes. The seat was adjusted to allow comfortable posture.
Figure 57. A test subject evaluating discomfort during the trial.
Vibration amplitudes were calculated using the standard’s frequency weightings and
multiplying factors, and r.m.s. averaging method. Point and OVTVs were calculated
based on the standard guidance.
6.3.6 Effect of sequence type and order to judgments
As the test subjects were exposed to each stimulus three times in random order, it was
necessary to validate that the judgment was not affected by the order of stimuli in a
sequence or the order of sequences. An averaged judgement value for each sequence
was calculated for the subjects based on the averaged values of each stimulus. The
values were then tested for any statistical differences using Friedman test and
Wilcoxon signed-rank test. The null hypothesis for Friedman test was that the
distributions are same across the repeated measurements. The null hypothesis was
rejected if p-value was low (p < 0.05). The null hypothesis for Wilcoxon test was that
148
the distributions of the samples are symmetrical. The null hypothesis was rejected if p-
value was low (p < 0.05).
6.3.7 Effect of subject characteristics to judgments
There have been indications in previous publications that men and women perceive
vibration differently (Mansfield 2005). The differences can be related to the differences
between physical or psychophysical characteristics (i.e. weight distribution, sensitivity,
etc). Even though the goal of the study was not to evaluate differences between men
and women, it was important to analyse any differences for improving the prediction of
discomfort. If no statistical differences are found between men and women, then also
the height or weight has most likely no effect to judgment, as clear differences in the
physical characteristics were noted between the genders.
Mann-Whitney U test was used to analyse if there were statistical differences between
the data samples. The test is used to show if two independent data sets come from the
same distribution, thus have no statistical differences. The null hypothesis in the test
was that the compared data samples come from identical populations, thus have same
distribution. P- and z-values were calculated for testing the null hypotheses. The p-
value was used to determine if the null hypothesis can be considered true.
6.4 Results
As rotational motion will always have translational components for the relevant axes,
the translational axes were present in all scenarios (Table 32). However, stimuli without
rotational components were well achieved. The table shows that in most cases the
preferred effect of the factors was achieved (compare to Table 30). The frequency
weighted r.m.s. values represented typical range of vibration levels in field
environments. Vibration was measured over all 9-axes even where there was no
driving signal, to include any cross-axis response of the seat dynamics. Highest
magnitudes were for stimuli C1, E1 and E3. Crest factors were below 9 for all stimuli
and axes.
149
2
Table 32. The frequency weighted r.m.s. values (m/s ) with respective multiplying factors of
each stimulus for each measured axis and overall vibration total value (OVTV) using seat
translational and rotational axes and backrest translational axes.
Seat translational Seat rotational Backrest translational Floor translational
Main
Variation x y z Roll Pitch Yaw x y z x y z OVTV
stimulus
A 1 0.308 0.284 0.702 0.224 0.136 0.029 0.580 0.168 0.251 0.080 0.086 0.259 1.080
A 2 0.192 0.162 0.701 0.117 0.077 0.026 0.419 0.095 0.246 0.063 0.063 0.257 0.906
A 3 0.154 0.107 0.704 0.062 0.059 0.026 0.387 0.059 0.242 0.062 0.059 0.258 0.867
A 4 0.429 0.401 0.706 0.318 0.197 0.030 0.756 0.242 0.255 0.100 0.108 0.260 1.296
A 5 0.404 0.215 0.297 0.165 0.185 0.018 0.686 0.135 0.135 0.074 0.057 0.106 0.931
A 6 0.149 0.108 0.273 0.054 0.037 0.012 0.220 0.056 0.116 0.059 0.047 0.100 0.422
B 1 0.250 0.803 0.349 0.452 0.128 0.052 0.511 0.434 0.078 0.088 0.237 0.146 1.230
B 2 0.163 0.705 0.158 0.381 0.084 0.041 0.355 0.354 0.043 0.053 0.205 0.065 0.979
B 3 0.082 0.659 0.133 0.308 0.043 0.023 0.199 0.316 0.034 0.047 0.190 0.060 0.836
B 4 0.270 0.554 0.765 0.374 0.139 0.047 0.564 0.338 0.149 0.123 0.171 0.347 1.259
B 5 0.153 0.272 0.718 0.185 0.080 0.026 0.390 0.167 0.137 0.081 0.086 0.328 0.925
B 6 0.103 0.199 0.728 0.091 0.055 0.017 0.365 0.103 0.136 0.082 0.083 0.329 0.869
C 1 0.741 0.803 1.197 0.528 0.381 0.073 1.303 0.472 0.251 0.248 0.149 0.380 2.250
C 2 0.400 0.417 1.126 0.287 0.197 0.043 0.874 0.244 0.232 0.148 0.094 0.364 1.615
C 3 0.302 0.238 1.135 0.126 0.122 0.031 0.792 0.116 0.223 0.150 0.078 0.358 1.471
C 4 0.335 0.389 0.458 0.262 0.157 0.030 0.625 0.227 0.102 0.093 0.068 0.142 1.010
C 5 0.171 0.208 0.432 0.138 0.076 0.018 0.382 0.123 0.093 0.056 0.050 0.133 0.674
C 6 0.123 0.130 0.437 0.064 0.038 0.011 0.308 0.064 0.090 0.060 0.045 0.132 0.580
D 1 0.472 0.435 0.331 0.361 0.233 0.058 0.883 0.284 0.092 0.236 0.106 0.100 1.257
D 2 0.225 0.323 0.297 0.161 0.084 0.012 0.375 0.158 0.083 0.198 0.089 0.091 0.671
D 3 0.111 0.171 0.257 0.081 0.043 0.008 0.225 0.085 0.068 0.090 0.052 0.078 0.423
D 4 0.473 0.246 0.240 0.194 0.224 0.052 0.845 0.164 0.077 0.256 0.068 0.056 1.086
D 5 0.351 0.163 0.287 0.083 0.124 0.011 0.479 0.078 0.084 0.319 0.049 0.067 0.705
D 6 0.341 0.087 0.277 0.050 0.118 0.010 0.460 0.040 0.077 0.310 0.032 0.059 0.661
E 1 0.620 0.915 1.461 0.498 0.286 0.070 1.009 0.565 0.239 0.477 0.523 0.668 2.258
E 2 0.349 0.461 1.371 0.255 0.146 0.039 0.647 0.274 0.201 0.300 0.281 0.638 1.687
E 3 0.211 0.210 1.389 0.117 0.093 0.028 0.599 0.119 0.181 0.188 0.144 0.634 1.566
E 4 0.296 0.147 0.317 0.082 0.094 0.014 0.336 0.086 0.081 0.229 0.056 0.103 0.594
E 5 0.660 0.216 0.243 0.098 0.353 0.022 1.205 0.123 0.098 0.221 0.103 0.077 1.468
E 6 0.352 0.550 0.286 0.406 0.105 0.030 0.473 0.403 0.082 0.241 0.094 0.075 1.040
150
Figures 58-60 show OVTVs using the seat translational and the rotational and the
backrest translational axes for each stimulus in a sequence. Most notable difference is
that sequence a had significantly higher amplitude for the first stimulus (E2) than
sequences b and c.
2
Figure 58. Sequence ‘a’ presented as the frequency weighted r.m.s. values (m/s ) using the
seat and the backrest translational and the seat rotational axes with the respective multiplying
factors.
2
Figure 59. Sequence ‘b’ presented as the frequency weighted r.m.s. values (m/s ) using the
factors.
151
2
Figure 60. Sequence ‘c’ presented as the frequency weighted r.m.s. values (m/s ) using the
factors.
6.4.2 Judgment style and averaging period
Figure 61 shows an example of a continuous judgment of a stimulus. The judgment

process was found to include three stages: 1) a delay for responding to vibration (2
seconds), 2) an adjustment period (3 seconds) and 3) a fine tuning period (10
seconds), where the final judgment was decided. A delay in average of 2 seconds was
noted for responses to vibration during the trial. This was also found out previously in a
pilot trial (Bhalchandra 2008). At the beginning of each stimulus also an adjustment
period was noted, at which time the subject settled for the initial judgment level. In the
last phase the subject changed the judgment relative to the initial level and vibration
exposure (this included the 2 second response delay as well).
152
Figure 61. An example of continuous judgment of vibration (lower figure shows acceleration of
seat vertical axis, above the discomfort judgment of the stimulus is shown).
Because of the response delay and the early adjustment period, first five seconds of
the judgment of each stimulus was not included in the analyses (Figure 61). The last 10
seconds were averaged to produce a single numerical value of subject’s discomfort
using arithmetic averaging. The scale of the judgment was normalised between 0 and 2
in 16 bit resolution (i.e. zero means “no discomfort” and two means “high discomfort”).
153
6.4.3 Judgment values
An averaged judgment value was calculated for each subject and sequence (Table 33).
The averaged value included all averaged judgments from the stimuli. The subjects
showed similar judgment levels for all sequences.
Table 33. Averaged judgment values for each sequence and test subject, and the order which
the sequences were run for each subject.
Stimuli sequences and mean judgments

Sequence
Test subject a b c
order (1,2,3)
1 0.982 0.895 0.896 a,b,c
2 0.443 0.481 0.512 b,c,a
3 0.547 0.578 0.516 c,a,b
4 0.762 1.062 1.019 a,c,b
5 0.814 0.974 1.011 c,b,a
6 0.802 0.821 0.813 b,a,c
7 0.737 0.761 0.667 a,b,c
8 0.790 0.997 0.930 b,c,a
9 0.699 0.746 0.701 c,a,b
10 0.698 0.780 0.931 a,c,b
11 1.056 1.210 0.947 c,b,a
12 0.384 0.461 0.397 b,a,c
13 0.508 0.482 0.456 a,b,c
14 0.859 0.755 0.900 b,c,a
15 0.651 0.659 0.642 c,a,b
16 1.143 1.171 1.287 a,c,b
17 0.662 0.646 0.875 c,b,a
18 0.422 0.462 0.366 b,a,c
19 0.388 0.681 0.779 a,b,c
20 0.586 0.489 0.628 b,c,a
21 0.398 0.492 0.365 c,a,b
22 0.631 0.886 0.857 a,c,b
154
Averaged judgments for each stimulus variation were also calculated for all subjects,
and for men and women separately (Figure 62). Men had slightly higher average
values for most stimulus variations, but the differences were small.
Figure 62. Averaged judgments for each stimulus for all subjects, and men and women
separately.
6.4.4 Effect of sequence type and order to judgments
The order of stimuli in each three sequences was randomised, so there was a
possibility that the judgment scale would change depending on sequence type and
order of the sequences. This can be validated by comparing the averaged judgments
for each sequence (Table 33).
For the effect of a sequence type to the judgment scale the Friedman Chi-Square value
was 6.636 and p-value 0.0362. The results suggest that the null hypothesis should be
rejected, thus there was a difference of judgment scale between sequence types.
Using Wilcoxon rank-test it was found that there were differences (p<0.05) between
sequences a and b (W=-151, z=-2.44, p=0.015). Between sequences a and c (W=-111,
z=-1.79, p=0.074) and b and c there were no differences (W=7, z=0.11, p=0.912),
although for a and c the p value was very close to 0.05.
For the effect of sequence order to the judgment scale the Friedman Chi-Square value
was 0.636 and p-value 0.7275. The results suggest that the null hypothesis is correct,
thus the order of sequences did not affect the judgments. This was confirmed using
Wilcoxon rank-test between orders 1 and 2 (W=-89, z=-1.44, p=0.150) and 1 and 3
(W=-43, z=-0.69, p=0.490), and between orders 2 and 3 (W=82, z=1.32, p=0.187).
155
6.4.5 Effect of subject characteristics to judgments
Visual inspection of the results showed that the men evaluated each stimulus
systematically slightly higher on average than the women (Figure 62). However in
many cases the difference was marginal and similar trend was shown. The judgments
of the men and the women were not significantly different (Table 34). Although the men
tended to score slightly higher on average, the difference was not significant (p>0.05).
Thus all further analyses were made without separating the subjects. The results also
suggested that weight and height did not affect the judgments.
Table 34. Mann-Whitney U test for differences in judgment between men and women.
n1 n2 U p (two-tailed)
30 30 381 0.315
Normal approx z = -1.02013 0.308
6.5 Discussion
Based on the field measurements it was found that the rotational and floor axes had
small contribution to OVTV, because of the standard’s weighting. The effect of the
different axes was needed to be validated in a multi-axis environment. Before validation
of the standard procedure is possible, it was necessary to validate the chosen
judgment method and effects of the study procedure to the results. This was a first
known study which used non-stationary and field-like multi-axis stimuli for evaluating
the correlation between discomfort and vibration exposure in a laboratory.
6.5.1 Vibration stimuli
The stimuli created for the laboratory trial was based on the field measurements in
Chapter 5. The purpose was to simulate similar vibration characteristics that are found
in typical work machines in the field. This was a different approach than in other
laboratory studies where sinusoidal or random vibration stimuli have been used. The
vibration amplitudes and frequency spectra showed typical vibration characteristics of
work machines, where frequencies less than 10 Hz are dominant. The rotational axes
were purposely emphasised for at least one of the main stimuli to evaluate if the
rotational axes have any effect on discomfort. However, the frequency weighting and
multiplying factors attenuated the rotational axes so that they were not dominant for
any stimulus when calculating OVTVs.
Crest factors were below 9 for all axes and stimuli. This was somewhat expected as
the test bench could produce vibration only between 1 and 20 Hz and the stimuli were
156
created by integrating acceleration data, which included bandpass filtering. However,
the magnitude levels of the stimuli were similar to the work machines they simulated.
6.5.2 Judgment style and values
The averaging using an arithmetic mean of the judgment values was based on the
results from noise research. It was noted that it took the subjects about 5 seconds to
find the initial discomfort level for each stimulus. For this reason the first 5 seconds
from each judgement data was discarded and last 10 seconds were averaged as a
single value.
The continuous judgment method allows more detailed analysis of the judgment style
and reaction to vibration than other normally used methods, as the judgment is
recorded at the same frequency as the stimulus.
6.5.3 Effects of procedure and method for results
There was an indication that the sequence types affected judgment scaling, where the
averaged judgments of sequence a had statistical differences compared to the other
sequence types (especially for sequence b). A closer look revealed that sequence a
had higher vibration magnitude for the first stimulus than sequences b and c. It is likely
that the first stimulus of a sequence influenced the scale for the judgments of rest of
the stimuli, because no absolute categories were given to the subjects. However, the
differences in the judgment values were small and because all subjects were exposed
to all sequences, no normalisation was applied. There was no indication that the order
of sequences affected the judgments.
There were no significant differences in the judgments between the men and women.
This suggests that neither gender nor physical dimensions affected the discomfort
judgment.
6.5.4 Scope of the results
The shaker used in the trial could produce movement between 0 and 20 Hz. In this
case, because of the high pass filtering, the frequency range was in practice between 1
and 20 Hz. However as most of the energy in mobile work machines are within this
frequency range, the shaker could be used to simulate vibration present in the field.
Even though the study design and shaker caused limitations to the usability of the data,
they were not considered critical considering the goals.
157
The standardised method includes twelve axes, but in this trial the floor axes were not
used in the analyses, thus only nine axes were analysed. This was because the
subjects kept their feet on a non-moving platform. This was done in order to reduce the
number of variables in this laboratory study. It was reasonable to assume that feet did
not make any significant contribution to the judgment (Bhalchandra 2008). However
there might be a change in correlation between vibration and discomfort if feet would
be included as well. This should be verified in future studies.
There are inconsistent or non-existent results for time-dependency between vibration

and discomfort (see Chapter 2). Even though this is an important issue, it was not
considered in this study, as the goal was to validate the standard’s method, which does
not require time dependency for making relative judgments. The duration of a stimulus
used in this trial (15 seconds) was not long enough to evaluate time dependency. The
effect of the length of the study procedure was minimised by having small breaks
between the sequences. No statistical differences were found for the judgments of the
first sequence to the last sequence.
The trial was limited to using subjects in a seated posture and leaning against a
backrest using a European small-car seat, which had previously been run-in. The
results might not accurately predict responses in a work machine seat, which has a
different design and results in occupants adopting a different posture. However stimuli
from the mobile work machines were used in the study, along with those from other
road and off-road vehicles.
As the trial was conducted in a laboratory, the subjects did not see any road or could
not predict or change the forthcoming movement. It has been shown that additional
information (e.g. noise and visual sight) can change the response of a single variable
(e.g. vibration) (Sato, et al 2007). Thus, the results might not be similar in a field
environment where other factors have more presence.
158
6.6 Conclusions
A laboratory trial was conducted to provide data for validating the standardised method
in a multi-axis environment. It was verified that the continuous judgment method can be
used to evaluate discomfort of subjects to vibration amplitudes:
• The effect of stimuli and sequence order was minimised by changing the order
of sequences for each subject and repeating each stimuli in random order for
three times for each subject;
• Small effects of stimuli order were found, but they did not affect the overall
results as judgment values were averaged;
• No statistical differences were found for the gender and physical characteristics;
• When using continuous judgment method, the first 5 seconds from the
judgment values should be neglected, because of the adjustment period and
delay of response.
Based on the validation of the data it was concluded that it can be used for analysing
the correlation between vibration and discomfort and to validate and optimise the ISO
2631-1 method.
159
7 Chapter 7 - Laboratory trial part II:
validating the ISO 2631-1 method
7.1 Introduction
The method in ISO 2631-1 can be used to evaluate and assess discomfort from whole-
body vibration. However, the standard includes very little information on the
assessment and guidance on how to interpret the results, or even how to choose the
proper axes. The standard allows different interpretations, because it does not explicitly
define the use of certain combinations of axes. However, it is strongly implied that at
least the seat translational axes should be used in the analyses. Depending on
circumstances, the backrest and floor axes should be included in the analyses. In
addition there is a possible need to include the rotational axes from the seat as well.
The full 12-axis method requires complex equipment which is not available to most
practitioners. It would be beneficial if as few axes as possible could be used for the
assessments, as this reduces complexity and cost, but currently there are no estimates
of how much this reduced data set compromises the accuracy of the discomfort
assessment. In many cases this has led to studies where only the seat translational
axes have been measured.
The results from previous publications have indicated that the standard ISO 2631-1
method might not be valid. Evaluations of vibration are affected by the axes measured,
frequency weighting curves, the multiplying factors and the averaging method. The
frequency weighting models the response of the body in the frequency domain and the
multiplying factors define the relative importance between the different axes. The
averaging method emphasises the frequency characteristics of the vibration (i.e. effect
of shocks compared to an average amplitude). Even though these methods have
significant effects on the evaluation, they have not been validated in a practical multi-
axis environment. It has been suggested that the current method does not provide
accurate results which are comparable between environments (Mansfield and Maeda
2005b). It is necessary to investigate their role to understand how the standard method
could be improved.
The frequency weighting curves have had considerable research and validation since
late 1960’s. The results show that the current frequency weightings for a seated person
(Wd and Wk) have been successful in predicting both human body transmissibility and
subjective response. The other frequency weighting curves (Wc and We) have not been
160
subject to such thorough validation, but currently there are no conclusive results to
support changing them. As the current instrumentation supports the weighting filters
and they have an evidence base, there is no reason to prioritise changing them at this
point. Thus the only practical option is to use them or not use them (i.e. no new
weightings should be proposed).
The multiplying factors for the axes have had less research and validation than the
frequency weighting curves despite their potential large modifying effects on vibration
values. The purpose of the factors is to emphasise different axes and measurement
locations. However there have been only few studies that have measured all twelve
axes and considered their effects on discomfort (Griffin 1990, Maeda 2004). The
current factors in ISO 2631-1 have no direct reference, thus the scope of the factors
are not referenceable in the literature. It can be assumed that the multiplying factors
are based on the results from Griffin et al. (1982), where it was found that comfort
contours for each axis had different magnitude level (See Figure 8 in Chapter 2).
However, the author has not found a publication where this has been replicated.
Because it is easy to apply a different set of multiplying factors without changing the
instrumentation, the validation of the factors should be considered.
The main averaging method to evaluate discomfort is to use the frequency weighted
r.m.s. values, but also additional methods exist. ISO 2631-1 refers to the VDV method
if shocks have a dominant effect (i.e. high crest factor). The VDV method is not
convenient to use when results are to be compared to r.m.s. values, as the values are
in different dimensions, with VDV increasing with the exposure time. The effect of
higher emphasis on shocks is easier to compare using the r.m.q. method suggested by
BS 6841 standard. The r.m.q. method gives the same emphasis to shocks as VDV, but
is in same format as the r.m.s. method. As the crest factors in the field measurements
showed values exceeding 9 (Chapter 5), the standard implies that higher emphasis on
shocks might improve correlation to discomfort. There have been some studies that
have compared differences of results in field using either the r.m.s. or r.m.q. methods
(Wikstroem, et al 1991). However no conclusive evidence of either method being better
was shown, thus there is a need to improve understanding the effects of shocks to
discomfort.
The results from the field measurements in Chapter 5 showed that the rotational and
floor axes were normally the least significant, because of the standard’s weighting and
multiplication factors. This suggests that basically the same overall vibration total value
161
(OVTV) can be achieved with less than twelve axes in most environments. Based on
the results, the following hypotheses were considered:
• Only a few (or one) most dominating axes will determine the subjective feeling;
o Other axes will have negligible effect and need not be measured and
analysed;
• The seat rotational and floor translational axes, with the vibration magnitudes
seen in the field measurements, have no practical effect on subjective feeling;
• OVTV based on the seat translational axes (including 1.4 factors) will largely
determine the discomfort level.
Based on the previous studies and literature the following hypotheses can be added:
• Frequency weighting improves the correlation between measured vibration and

discomfort;
• Multiplying factors should be larger for the horizontal axes than other axes;
• Higher emphasis to shocks (i.e. r.m.q. and VDV) improves the correlation;
• Inclusion of axes additional to the seat translational will improve the correlation.
There is a need to better understand which axes and locations contribute to the
discomfort judgment and thus correlate with it. Also it is possible to optimise the
number of locations and axes needed to be measured to obtain practical accuracy. For
efficiency and ecological validity it is important to validate the standardised method
based on stimuli that is present in real environments, because that is where people are
exposed to it.
7.2 Goals of the trial part II

The goals of part II of the laboratory study are to validate the standard method and to
evaluate the roles of frequency weighting, multiplying factors and averaging methods
for predicting discomfort. If the standard is correct on emphasising the axes, then the
evaluation process can be significantly simplified, as the rotational and floor axes are
not needed. The analyses aim to answer the following questions:
• Does the correlation between vibration and discomfort improve when additional
axes are included?
162
• Are there practical differences in correlation using different combinations of
axes?
• Are the rotational axes as insignificant as the standard method implies based
on the field measurements?
• Does the frequency weighting and the multiplying factors improve the
correlation?
• Will more emphasis on shocks (i.e. the r.m.q. method) provide any better
correlation than the default method (i.e. the r.m.s. method)?
7.3 Methods
The details of the laboratory trial, stimuli and subjects are presented in Chapter 6.
Vibration magnitudes were calculated for the frequency weighted and unweighted
acceleration data with and without the multiplying factors and using the r.m.s. and
r.m.q. averaging methods. Thus in total six sets of results for all axes, PVTVs and
OVTVs were produced to compare the effects of weighting and averaging methods
(see Table 7 in Chapter 3).
7.3.2 Selecting combinations of axes for calculating overall vibration

total values
The standard allows several possible combinations of axes to be used in the analyses.
Only three translational axes from the seat are required to be measured, but in a
minimum case only one axis is enough to conduct the evaluation (if other axes are less
than 25% of the dominant axis). Based on the guidance in ISO 2631-1 the following
scenarios to produce OVTV are allowable for evaluating discomfort of whole-body
vibration (for seated persons):
• A dominant axis (which could be any of the twelve axes);
• A dominant PVTV (rest of the PVTVs are less than 25 % of the dominant
PVTV);
• A combination of PVTVs of seat, backrest and floor (where PVTVs are larger
than 25 % of the dominant PVTV);
• All twelve axes.
163
Because of the number of combinations which are possible, the following practically
realisable scenarios were used to analyse correlation between discomfort and vibration
(Table 35).
Table 35. Practically realisable scenarios for calculating an overall vibration total value based on
the ISO 2631-1 standard when using the seat, backrest and floor translational axes.
Scenario
Explanation
number
Overall vibration total value of seat translational axes (without 1.4 multiplying factors for
1
horizontal axes)
Overall vibration total value of seat translational axes (with 1.4 multiplying factors for
2
horizontal axes)
Overall vibration total value based on point vibration total values of seat translational axes
3
(without 1.4 multiplying factors) and backrest fore-and-aft axis (with 0.8 multiplying factor)
4
(without 1.4 multiplying factors) and backrest translational axes (with multiplying factors)
Overall vibration total value based on point vibration total values of seat translational and
5 rotational axes (without 1.4 multiplying factor for seat horizontal axes, but with multiplying
factors for rotational axes)
6 rotational axes (with 1.4 multiplying factor for seat horizontal axes and with multiplying factors
for rotational axes)
7 rotational axes and backrest translational axes (without 1.4 multiplying factor for seat
horizontal axes and with all multiplying factors for other axes)
The scenarios were chosen for practical purposes whilst including most possible
interpretations of ISO 2631-1. The floor axes were not used in any scenarios as the
laboratory trial design did not include the effects of the floor vibration in order to limit
the number of variables.
It would be feasible to greatly simplify the vibration measurement, if less than twelve
axes are needed in the analyses without losing the accuracy. Normally the whole-body
vibration measurement equipment include only the standard 3-axis seat pad sensor.
Thus Scenarios 1 or 2 would be the most preferable.
Some commercial whole-body vibration measurement equipment include also a fourth

sensor channel, and so Scenario 3 would be possible. However, the standard and field
results imply that, in addition to seat translational axes, next importance is to measure
all backrest axes, thus Scenario 4 is important to consider.
Because the effect of the backrest axes to discomfort can be simulated using the 1.4
multiplying factors, it could be possible to measure the translational and rotational axes
from the seat and use the factors instead of measuring the backrest. Comparing
Scenarios 1 to 2, and Scenarios 5 to 6, can be used to predict if the backrest axes
need to be measured or if it can be replaced by the multiplication factors.
164
Analyses without rotational axes are the most attractive option, as they are hardest to
measure in practice (and are potentially the most erroneous). This can be evaluated
comparing the correlation of the scenarios with and without the rotational axes.
The effect of all possible axes can be evaluated based on Scenario 7.
7.3.3 Judgment method
The discomfort of subjects was acquired using the continuous judgment, cross-modal
matching method. Arithmetic mean of last 10 seconds of judgment of each stimulus
was calculated based on the analysis in Chapter 6.
7.3.4 Correlation analyses
The goal of the analyses was to find correlation between discomfort judgment and
calculated vibration magnitudes of the chosen scenarios and different setups for
calculating vibration magnitudes. Thus statistical analysis of coefficient of correlation
was required. In this case Spearman method was used (r2). Additionally Kendall’s tau
was calculated to validate the correlation results. Wilcoxon signed-rank test was used
to evaluate any statistical differences amongst the results.
7.4 Results
The correlation values using Kendall’s tau produced systematically similar results than
Spearman rho, thus only Spearman values are shown.
7.4.1 Correlation between vibration magnitudes and judgments
Twenty-two test subjects evaluated 30 different stimuli three times in random order (the
procedure explained in detail in Chapter 6). The data set was compared to the seven
scenarios to determine the correlation for vibration and discomfort. Mean judgment
values of the vibration generally increased with OVTVs of each stimulus in a nominally
linear fashion (Figure 63). The correlation tended to improve when more axes were
included in the analysis, the best occurring for the full 9-axis analysis (Scenario 7, r2
0.850). However, it was also evident that the correlation (r2) was almost identical for
Scenarios 2 (0.823), 4 (0.836) and 6 (0.844). Scenario 1 had the worst correlation (r2)
of 0.623. The correlation (r2) of Scenario 3 (0.799) and Scenario 5 (0.743) were better
than for Scenario 1, but were clearly worse than the best scenario. The correlation
between discomfort and vibration was positive and linear.
165
Figure 63. Correlation between mean judgements of all subjects and frequency weighted
vibration magnitudes using the r.m.s. method and respective multiplying factors for chosen
scenarios.
The correlation value changed significantly for each test subject (Table 36). In half of
the cases (11/22) OVTV calculated using Scenario 6 provided the best correlation.
Scenario 7 provided best correlation in 10 cases and Scenario 3 in one case. The best
average correlation was for Scenario 7, but there was no statistical difference
compared to Scenario 6 using Wilcoxon signed-rank test (p=0.617). Also Scenario 2
and 7 showed little statistical difference between correlations (p=0.0735). However, a
166
clear statistical difference was found between Scenarios 1 and 2 (p<0.001); thus using
the 1.4 multiplying factors for seat horizontal axes improved the correlation. Differences
between Scenarios 1 and 7 (p<0.001) showed that additional axes improved
correlation.
2
Table 36. The correlation (r ) between discomfort judgments of individual subjects and
frequency weighted vibration magnitudes using the r.m.s. method with respective multiplying
factors for each scenario (the best scenario for each subject is embolden and underlined).
Scenarios
5
1 2 3 4 6 7
Seat
Subject Seat Seat trans+1.4 Seat Seat +back Seat trans+rot+1.4 Seat trans+rot, back
trans+rot
trans factors trans+back x trans factors trans
1 0.304 0.408 0.335 0.373 0.358 0.420 0.383
2 0.499 0.606 0.536 0.567 0.570 0.611 0.584
3 0.473 0.611 0.520 0.555 0.551 0.612 0.555
4 0.417 0.569 0.610 0.624 0.530 0.587 0.635
5 0.507 0.682 0.632 0.671 0.614 0.694 0.692
6 0.437 0.575 0.486 0.532 0.506 0.585 0.550
7 0.343 0.444 0.431 0.465 0.415 0.454 0.468
8 0.500 0.697 0.642 0.676 0.615 0.710 0.699
9 0.269 0.392 0.301 0.329 0.330 0.396 0.342
10 0.374 0.440 0.518 0.515 0.405 0.456 0.516
11 0.402 0.503 0.522 0.540 0.468 0.515 0.552
12 0.338 0.532 0.555 0.573 0.452 0.554 0.586
13 0.419 0.627 0.560 0.589 0.536 0.645 0.604
14 0.552 0.676 0.697 0.730 0.633 0.686 0.731
15 0.482 0.591 0.612 0.629 0.534 0.611 0.642
16 0.436 0.610 0.539 0.579 0.553 0.635 0.602
17 0.483 0.632 0.614 0.640 0.597 0.653 0.644
18 0.457 0.564 0.559 0.580 0.548 0.569 0.597
19 0.473 0.648 0.577 0.612 0.586 0.664 0.626
20 0.345 0.470 0.464 0.485 0.415 0.480 0.496
21 0.431 0.521 0.606 0.610 0.500 0.541 0.614
22 0.311 0.422 0.449 0.464 0.378 0.428 0.485
Average 0.421 0.555 0.535 0.561 0.504 0.568 0.573
167
The results show that using a single location will not improve correlation compared to
multiple locations (Table 37). The best correlation was for the backrest, but also the
dominant location for each stimuli showed practically similar correlation (p=0.842). The
seat translational axes showed clearly worse correlation compared to the backrest
axes (p=0.0001). Also the seat rotational axes showed statistically lower correlation
than the backrest axes (p=0.007).
2
Table 37. The correlation (r ) between discomfort judgments and frequency weighted point
vibration total values with respective multiplying factors and the dominant (i.e. largest) point
vibration total value for each test subject (best correlations are embolden and underlined).
2
Point vibration total values (m/s )
Seat Back
Subject Dom Seat Rot
Trans Trans
1 0.340 0.304 0.370 0.356
2 0.548 0.499 0.466 0.493
3 0.527 0.473 0.429 0.410
4 0.557 0.417 0.500 0.614
5 0.631 0.507 0.477 0.581
6 0.478 0.437 0.458 0.485
7 0.398 0.343 0.494 0.476
8 0.601 0.500 0.585 0.653
9 0.333 0.269 0.354 0.304

10 0.487 0.374 0.271 0.486
11 0.508 0.402 0.362 0.489
12 0.497 0.338 0.489 0.603
13 0.564 0.419 0.474 0.485

14 0.627 0.552 0.462 0.626
15 0.595 0.482 0.467 0.608
16 0.535 0.436 0.607 0.562
17 0.588 0.483 0.635 0.592
18 0.546 0.457 0.382 0.498
19 0.542 0.473 0.553 0.519
20 0.445 0.345 0.354 0.422
21 0.570 0.431 0.362 0.532
22 0.400 0.311 0.403 0.536
Mean 0.514 0.421 0.453 0.515
168
7.4.2 Effect of weighting and averaging methods
Applying the frequency weightings improved the correlation for all scenarios, except for
Scenario 1 (Table 38). The scenario of the best correlation changed to 4 without the
frequency weighting, but the best overall correlation remained Scenario 7 with the
frequency weighting. The results from Scenario 1 and 2 show that weighting improved
correlation when horizontal axes have more effect, thus suggesting that the vertical
weighting is less effective than the horizontal weighting.
Scenarios from 3 to 7 included either the backrest or the rotational axes, thus their
results were affected by the multiplying factors (Table 38). The results show that the
correlation improved in all cases where multiplying factors were not used. Scenario 7
had the best correlation in both cases (with and without the factors). Effect of 1.4
multiplying factor was evaluated by comparing the correlation between Scenarios 1 and
2 to Scenario 4, which included the backrest axes. The results show that in each case
where 1.4 multiplying factors were used the correlation was better than using the
backrest axes.
The results show that none of the scenarios had better correlation with the r.m.q.
method. Although the differences were small, they were systematic.
2
Table 38. Comparison of correlations (r ) for using the r.m.s. and r.m.q. averaging methods for
the frequency weighting with the multiplying factors, the frequency weighting without the
multiplying factors and without the frequency weighting and the multiplying factors.
OVTV Scenarios
3 5
2 4 6 7
1 Seat Seat
Seat Seat Seat Seat
Seat trans+ trans+r
trans+1.4 +back trans+rot+ trans+rot,
trans back ot
factors trans 1.4 factors back trans
x
r.m.s. Weighting
0.417 0.551 0.530 0.557 0.500 0.564 0.569
values w multp
Weighting.
0.417 0.417 0.538 0.578 0.573 0.586 0.591
w/o multp
W/o all 0.460 0.460 0.509 0.524 0.443 0.446 0.466
r.m.q. Weighting
0.376 0.516 0.523 0.530 0.442 0.534 0.551
values w multp
Weighting.
0.376 0.376 0.527 0.545 0.541 0.566 0.569
w/o multp
W/o all 0.426 0.426 0.480 0.474 0.436 0.430 0.435
169
7.5 Discussion
7.5.1 Correlation between vibration and discomfort
The correlation calculated from the mean judgments of all test subjects was better than
the mean correlation of each subject. This was an expected result, as averaging
reduces the influence of outliers. A high correlation between the averaged discomfort
values and vibration was found (r2= 0.850; scenario 7). Based on the interpretation of
the Spearman r2 value, the result indicates that the vibration exposure alone
contributed 85 % of the change in discomfort judgment(. The best correlation (r2) for
individual test subjects varied between 0.396 and 0.731. Thus there was a large
variability between the subjects. However, for each subject the trend was similar:
additional axes improved correlation. The correlation to vibration was positive and
linear.
The results showed that the best correlation, using the standard’s weighting procedure,
was Scenario 7, which included the translational and rotational axes from the seat
surface and the backrest axes. However, Scenarios 2 and 6, which used 1.4 factors to
replace the backrest axes, had practically the same correlation as Scenario 7. Thus 1.4
factors can be used to replace the backrest axes in most cases with minimal reduction
in predictive power for discomfort. Scenario 1 (seat translational axes), which would be
convenient for the current commercial measurement equipment, showed the lowest
correlation of all scenarios. It is clear that the additional axes improve correlation, or at
least the 1.4 factors should be used, if translational seat axes are only measured. In
many cases studies have used Scenario 1 for discomfort evaluation instead of scenario
2, which is the recommended way. For those data sets using Scenario 2 would be an
easy option as it does not require additional data.
The field measurements implied that for most environments the rotational axes have
only a small contribution to OVTV. This is because of small multiplying factors and the
effect of the frequency weighting curves. The results in this study indicated that
although the seat rotational axes did improve correlation, the effect was again relatively
small (Scenario 4 compared to Scenario 7).
7.5.2 Effect of frequency weighting
The weighting curves improved correlation in practically all cases, but the effect was
not significant. Vibration characteristics from a seat of a large work machine include
mainly low frequencies, because of low dynamic damping factors of tires, rubber
170
bushings, and seat. Thus the frequency weighting curves of the standard do not
significantly change the frequency content. Because the laboratory trial used stimuli
from field measurements, the effect of frequency weighting was small. The weighting
would have a larger effect when using either flat spectrums or sinusoidal vibrations.
7.5.3 Effect of multiplying factors
There was no evidence that the standard’s multiplying factors for the rotational and
backrest axes improved correlation. This was seen for all scenarios and PVTVs. In fact,
the results showed that correlation was worse when using the multiplying factors. All
PVTVs showed better correlation without multiplying factors. However, for the best
scenario, the correlation only marginally changed with or without the multiplying factors,
which implies that the multiplying factors improve correlation for some individual axes.
It depends on the combination of axes if the factors are useful or not. There might be
an alternative set of multiplying factors which would be more optimal and improve the
correlation.
The stimuli of the laboratory trial can be used to analyse the effects of the multiplying
factors for the seat horizontal axes. It was previously concluded that Scenario 2 was
better than Scenario 1. The difference between the scenarios is the 1.4 multiplying
factors for the seat horizontal axes (Scenario 2), thus any improvement in correlation is
caused by the factors. A previous study (Maeda and Mansfield 2006b) also indicated
that higher factors for horizontal axes will improve the correlation.
By analysing each stimuli type used in the trial, it can be indicated if the factors
improved correlation for all types of stimuli or just for some. Figures 64 to 68 show how
Scenarios 1 and 2 affect the correlation for each main stimuli. Scenario 2 was
systematically better, but differences varied significantly. Stimuli types A, B and E were
significantly better for scenario 2 than 1. Scenarios C and D showed almost similar
correlations. Detailed analyses of the frequency spectra of the main stimuli (Figure 50
in Chapter 6) showed no clear reason for the differences in the main stimuli. The same
conclusions were made also based on the vibration magnitudes (Table 32 in Chapter
6).
171
Figure 64. Differences in correlation for stimuli type A for scenarios 1 (left) and 2 (right) using
averaged judgments from all subjects.
Figure 65. Differences in correlation for stimuli type B for scenarios 1 (left) and 2 (right) using
Figure 66. Differences in correlation for stimuli type C for scenarios 1 (left) and 2 (right) using
Figure 67. Differences in correlation for stimuli type D for scenarios 1 (left) and 2 (right) using
172
Figure 68. Differences in correlation for stimuli type E for scenarios 1 (left) and 2 (right) using
Figure 69 and Figure 70 show all stimuli types for Scenarios 1 and 2. This verifies the
overall effect of 1.4 multiplying factors for improving the correlation. Based on the
results it was evident that the fore-and-aft and lateral axes should be more emphasised
compared to the vertical axis.
Figure 69. Correlation of averaged judgments and stimuli types using scenario 1 (seat
translational axes with 1.0 multiplying factors for all axes).
Figure 70. Correlation of averaged judgments and stimuli types using scenario 2 (seat horizontal
axes with 1.4 multiplying factors).
173
7.5.4 Effect of emphasising shocks
The correlation did not improve when shocks were emphasised. This was analysed
comparing correlations between the r.m.s. and r.m.q. methods. BS 6841 standard
suggests using the r.m.q. method instead of the r.m.s. method in situations where
shocks are expected to influence the results. The r.m.q. method showed a
systematically poorer correlation than r.m.s. for all stimuli, thus there was no evidence
that the use of vibration dose value (VDV) would improve correlation for the types of
stimuli used in the trial. However, it should be noted that the stimuli used in this study
were not designed to expose subjects to large shocks and so results from the tests in
more severe environments might show the opposite trend. However, it is interesting to
note that the r.m.q. method has been found worse also when shocks are present
(Wikstroem, et al 1991), however, overall no clear differences have been found
between the methods in most environments (Mansfield, et al 2000).
7.5.5 Optimal usage of the standard method
Based on the results, it is possible to improve the ISO 2631-1 method. The multiplying
factors degraded correlation for most axes and scenarios. The results hinted that
different factors for the additional axes will improve correlation. Also the 1.4 multiplying
factor for the seat horizontal axes implied that the factors for horizontal axes should be
higher compared to the vertical axes.
Individual PVTVs correlated inferior to OVTVs of any scenario. The results showed that
dominant PVTV is not enough to evaluate discomfort. Using a single axis (even if it is
dominant) is also problematic. In some cases, there might be no vibration on the axis.
However, if one suggests that, for example, only a vertical axis will correlate with
discomfort, then it should do it in all cases, not just in cases where the axis is clearly
dominant. In any case, using the dominant single axis or PVTV for discomfort
evaluation is not a good solution.
The optimal setup was found to be measuring the seat translational axes and
evaluating them using the frequency weighted r.m.s. averaging with 1.4 multiplying
factors for the horizontal axes. The correlation was practically the same as when
including all axes, but the measurement configuration is much simpler.
174
7.5.6 Comparison to previous studies
Compared to other studies found in the literature review (Chapter 2), similar and new
results were found. In this case the r.m.s. method was better than the r.m.q. method,
which has been also concluded from a field study (Wikstroem, et al 1991). Effect of
frequency weighting was concluded to improve correlation, but only slightly. This was
also concluded by Mansfield et al. (2000). Els (2005) concluded that there was no
difference between frequency weighted or unweighted r.m.s. values to the discomfort.
The results showed similar results to Parsons and Griffin (1983), that including more
axes will improve correlation.
7.5.7 Scope and limitations of the results
The vibration amplitudes from the backrest depend on the location of the sensor, thus
the correlation might be different if it was placed on a different location. However,
based on the standard guidance it is hard to determine the best location as backrests
have different height and there is no specific information on the position of the sensor.
Thus the argument of using the 1.4 multiplying factors to replace the backrest axes is
valid and potentially more repeatable, as the results give almost similar responses.
The study was limited to using subjects in a seated posture and leaning against the
backrest using a European small-car seat, which had previously been run-in. For this
type of seat, the vibration at the seat cushion and backrest has been transmitted
through the seat mounting points and therefore would be expected to correlate. The
results might not accurately predict responses in a work machine seat, which has a
different design and results in occupants adopting a different posture with different
backrest contact. However, the stimuli from mobile work machines were used in the
study, along with those from other road and off-road vehicles.
Most studies, which have investigated the evaluation method of ISO 2631-1, have used
sinusoidal or random stimuli (see Chapter 2). This study used stimuli which were
adapted from those measured in the field (Chapter 5) and thus were more
representative of real exposure situations. As there are no similar studies, it is not
possible to evaluate how the results compare regarding other types of field-like stimuli.
Although the amplitude ranges of the stimuli were limited, they covered a wide variety
of vibration exposure characteristics that commuting people are exposed in every day
work life, such as cars, buses and trains.
175
7.5.8 Further work
A limited number of stimuli and subjects were used, thus further work should be
conducted in the laboratory and field to verify the results of this trial. It is recommended
to use field-like stimuli of other environments for different durations and conduct similar
types of analyses for correlation. Additionally it is suggested that a different method
(e.g. the Magnitude Estimation method) is used, instead of the continuous judgment
method.
7.6 Conclusions
The results showed that a significant correlation between vibration exposures and
discomfort can be achieved using the ISO 2631-1 method in the laboratory. The
ambiguous guidance of the standard allows wide variety of possibilities to combine
axes and locations, but greater complexity in analysis does not necessarily guarantee a
better correlation between the measured vibration magnitude and ratings of discomfort.
For validating the standard procedure and finding the best setup, the following
conclusions were made:
• The correlation between vibration exposure and discomfort was best when all
possible axes were included; however:
o small statistical differences were found when using just seat translational
axes with 1.4 multiplying factors instead of the full 12-axis method;
o rotational axes had marginal effect;
• The use of frequency weighting only slightly improved the correlation, which is
assumed to be caused by the low frequency content of the stimuli;
• The multiplying factors of the backrest and rotational axes degraded correlation;
• The r.m.s. method provided slightly, but systematically, better correlation than
the r.m.q. method;
• No explanation was found why the 1.4 multiplying factors seemed to be better
suited for some stimuli characteristics than others;
o However, the factors still improved correlation for all stimuli types.
In the light of the results and the complexity of the current standard, it can be
concluded that there is a need and possibility to improve and simplify the current
method, or to develop a completely new method. The primary goal of the development
176
is to improve the current standardised method by optimising the measurement process
(i.e. minimising the number of measured axes) and the analysis procedure. The
multiplying factors are the best approach to improve the correlation.
177
8 Chapter 8 - Laboratory trial part III:
improving the ISO 2631-1 method
8.1 Introduction
The results from Chapter 7 showed that the correlation between the r.m.s. values and
discomfort judgment changed depending on how the weighting and multiplying factors
were used and which axes were included. The results showed that replacing the
backrest axes by multiplying the seat horizontal axes with 1.4 factors gave practically
the same correlation as including the backrest axes. Also it was evident that the
standard’s multiplying factors for the backrest and rotational axes decreased the
correlation, even though their purpose is to improve it. Although the rotational axes
improved correlation overall (even when using the multiplying factors), the effect was
marginal.
The conclusions from Chapter 7 suggested that by optimising the multiplying factors
the correlation between vibration exposure metrics and discomfort can be improved by
the simplest and most effective way. However, because of a large number of possible
combinations of factors for all twelve axes, a reduction of factor ranges need to be
done before any calculations. This can be done based on the results from the previous
chapters which showed that:
• Using 1.4 multiplying factors (instead of 1.0) for the seat fore-and-aft and lateral
axes indicated improvement in correlation;
• The multiplying factor for the backrest fore-and-aft axis improved correlation;
• The multiplying factors for the seat rotational axes and the backrest horizontal
axes degraded correlation;
• The floor translational and seat rotational axes had marginal contribution to the
overall vibration total value (OVTV) and discomfort;
• Most significant axes were the seat translational axes and the backrest fore-
and-aft axis.
It seems probable that by emphasising the seat horizontal axes, the correlation would
improve, because the 1.4 factor had a positive impact. This suggests that trying higher
factors for the seat horizontal axes would be a good starting point to improve
correlation. Another study has shown also problems relating to the multiplying factors
178
(Maeda and Mansfield 2006b). The study indicated that the horizontal axes should be
more emphasised compared to the vertical axis. However, the stimuli were artificial (i.e.
frequency content did not exhibit characteristics of a work machine), thus a study to
confirm the findings using field-like stimuli should be conducted.
8.2 Goals of the trial part III

Based on the practical possibilities and the results from Chapter 7 several parts of the
standard method were considered to be adequate and not to be a subject of change at
this time:
• Averaging vibration values using the r.m.s. method;
• The location of the measurement points and the axes;
• The root-sum-of-squares (r.s.s.) for combining the r.m.s. values of each axis
and location;
• The frequency weighting curves.
The goal is to find best possible combination of multiplying factors for each of the
measured nine axes so that the results correlate best with the discomfort judgment.
The results should show what relative emphasis of the axes will give the best possible
correlation. As the optimisation of the multiplying factors seems to be the most effective
way to increase correlation, two different methods should be used to find and validate
best multiplying factors for each axis: 1) brute force and 2) multiple linear regression
model.
8.3 Methods
The trial procedure, subjects and judgment method are explained in detail in Chapter 6.
The calculation of vibration magnitudes, correlation methods and standard scenarios
are explained in detail in Chapter 7.
8.3.1 Finding optimal combination of multiplying factors
8.3.1.1 Brute force
Brute force search is a trivial, but very general problem-solving technique. The
algorithm tests all possible combinations without any optimisation. The brute force
algorithm is simple to implement and will always find the answer if it exists. If the
combinations (i.e. calculation time) can be limited to a practical level, then the method
179
is the best option. The problems relating to other, more intelligent, methods are that the
implementation will take longer and more errors can occur in the analyses. An example
of an intelligent method is the Nelder-Mead method (also known as downhill simplex
method), which uses geometric shapes to search the optimum value without going
through all possible combinations. However the method requires pre-knowledge of the
problem and theory to be implemented and does not guarantee a result. A further
problem with the Nelder-Mead method is that optimisation will only find a single best-fit
for each set of initial conditions and therefore multiple searches with many sets of initial
conditions are required, with no guarantee of finding the best optimisation for complex
multi-factorial systems.
To use the brute force search method the first task is to optimise the number of
possible combinations. Using all nine measured axes and a range of multiplying factors
from 0.0 to 3.0 (best guess based on the previous results) at 0.1 resolution gives
26439622160000 possible combinations (319) to test. If an average computer can
calculate and test 1000 combinations in a second, then it will take 306014 days to
finish, which is not acceptable.
Because the factors are relative to each other, the factor of one of the axes does not
need to change, thus limiting the combinations to 318, which can be calculated in 9871
days. Also by reducing the range and resolution of the factors, the combinations can be
significantly reduced. At the beginning it is not necessary to use 0.1 step resolution,
because the first goal is to find approximate relations between the factors. If 0.2 step
resolution is used, then the calculation time is reduced to only 30 days. So using 4-5
normal PC-computers the analysis can be conducted in a week.
Table 39 shows the range and steps of multiplying factors tried for each axis. The seat
vertical was considered the axis to which rest of the axes were compared to, thus the
results show relative emphasis to the vertical axis. As the seat translational axes were
identified as the most important axes, the resolution of the step was 0.1 for them. For
the rest the step size was 0.2. So the final number of combinations used was 1.70x109
for the initial brute force optimisation and it was shown that by using a Python
programming language running on a normal PC that 1000 calculations per second was
possible. The task was divided to four Pentium PC’s, thus reducing the calculation time
to less than 5 days.
180
Table 39. The range and steps of multiplying factors for brute force search.
Axis Range Step

Seat fore-and-aft [0.0...3.0] 0.1
Seat lateral [0.0...3.0] 0.1
Seat vertical [1.0] -
Seat roll [0.0...2.0] 0.2
Seat pitch [0.0...2.0] 0.2
Seat yaw [0.0...2.0] 0.2
Backrest fore-and-aft [0.0...2.0] 0.2
Backrest lateral [0.0...2.0] 0.2
Backrest vertical [0.0...2.0] 0.2
8.3.1.2 Multiple linear regression model
Another method to find covariance between vibration values and judgments is to

calculate a regression model. In this case we can use a multiple linear regression
model to calculate regression between all nine axes (i.e. independent variables) and
judgment value (dependent variable). The general purpose of the multiple regression is
to learn more about the relationship between several independent variables and a
dependent variable.
The selection of the independent variables and the dependent variable has following
general rules:
• The correlation between each independent variable to the dependent variable

has to be linear;
• The independent variables should not correlate with each other;
• All variables should have been measured in continuous scale (such as interval
or ratio scale);
• The number of observations should be at least five times higher than the
number of independent variables (at least ten times is recommended).
In this case the dependent variable was the averaged judgment value of each stimulus
and independent variables were the frequency weighted r.m.s. values of the measured
axes. However, because of the measurement setup, some axes significantly correlated
(e.g. the seat and backrest fore-and-aft). Only axes that did not correlate significantly
could be used in the regression model. A correlation test was made to the axes.
The normal approach using the linear regression modelling is to assume linear
summing of independent variables:
Yi = β 0 + β1 X i1 + β 2 X i 2 + ... + β n X in (25)
181
Where Yi is the dependent variable, β n the model coefficients and Xn the independent
variables.
Even though the linear regression model is based on the linear assumption the
independent variables do not need to conform to this (as long as the dependent Yi is
linear in the coefficients β n ). As the standard weights the r.m.s. values of each axis
using the second power summing, then this could be also done to the independent
variables of the regression model:
Yi = β 0 + β1 X i21 + β 2 X i22 + ... + β n X in2 (26)
Both approaches are valid and will be used to determine the multiplying factors.
8.3.2 Visual evaluation of results
Visual evaluation of the effect of changing multiplying factors was made for the seat
translational axes using a contour map, where the multiplying factor for one axis
remains fixed while the factors for the other axes increase. The contour map will show
the elevation of correlation using colour map in 2D.
8.3.3 Validation of results
Validation of the found multiplying factors was done by dividing the subjects into two
groups randomly. The correlations for both groups were calculated and compared
based on the results. The validation was also done for data from other publications to
verify if the results can be used to conclude the effect more generally.
8.4 Results
All analyses were made also using Kendall’s tau correlation and practically the same
results were obtained. Both methods showed systematically the same conclusions,
thus the results here are only presented using the Spearman method.
8.4.1 Best combination of multiplying factors
8.4.1.1 Brute force
The brute force optimisation of the multiplication factors showed that it was possible to
improve the correlation between the weighted vibration and the subjective responses
(Table 40). The results show that the best correlation found was better than using any
standard scenarios (Spearman r2 0.850). The best correlation was found using only the
182
seat translational axes with emphasising the fore-and-aft and lateral axis compared to
vertical. The best top ten correlations were achieved without the backrest or rotational
axes (the factors were zero). The same correlation was possible to achieve with
different combinations of multiplying factors. This suggests clustering of the best
values.
Table 40. The top ten combinations of multiplying factors for all measured nine axes.
Spearman Seat Back

2
r P x y z roll pitch yaw x y z
0.949920 < 0.001 2.7 1.8 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.946454 < 0.001 2.7 1.7 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.944723 < 0.001 2.6 1.6 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.944723 < 0.001 2.6 1.9 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.944723 < 0.001 2.7 1.9 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.942994 < 0.001 2.5 1.7 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.942994 < 0.001 2.6 1.7 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.942994 < 0.001 2.6 1.8 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.942994 < 0.001 2.8 1.7 1.0 0.0 0.0 0.0 0.0 0.0 0.0
0.942130 < 0.001 2.4 1.6 1.0 0.0 0.0 0.0 0.0 0.0 0.0
A more detailed brute force search using step resolution 0.01 produced somewhat
better correlation (r2 from 0.949920 to 0.96297) and showed a small contribution of the
additional axes. However, no significant differences were found and since in practice
the second search is too precise and the seat translational axes alone produced a
good correlation, the results from the smaller step resolution were not considered in the
analysis.
Based on the results an optimised model was created, providing the best set of
multiplying factors for the vibration stimuli used in the experiment:
( )
1
av = 2.7 2 a wx
2
+ 1.8 2 a wy
2
+ 1.0 2 a wz
2 2 (3)
where av is discomfort value, awx is the frequency weighted r.m.s. value for the seat
fore-and-aft axis, awy is the frequency weighted r.m.s. value for the seat lateral axis, awz
is the frequency weighted r.m.s. value for the seat vertical axis.
183
8.4.1.2 Linear regression models: selecting independent and dependent
variables
The results showed that there was a high correlation between axes of same direction
(Table 41). For example the seat fore-and-aft axis was highly correlated with the seat
pitch and backrest fore-and-aft axes. The correlation was significant (p < 0.005).
However, the correlation between the different directions was not significant. The
rotational axes were significantly correlated with the respective translational axes (e.g.
the seat lateral and roll), thus they could not be used. Also the backrest axes were
directly correlated with the respective seat axes. Based on the results only either the
seat or backrest translational axes could be used in the regression model. Based on
the practicality, the seat translational axes were chosen.
Table 41. The correlations (Pearson r) and p-values for all independent variables.
Pearson (r)
Seat x Seat y Seat z Seat roll Seat pitch Seat yaw Back x Back y Back z
Seat x 1.000 .370 .281 .466 .971 .608 .933 .450 .316
Seat y .370 1.000 .258 .956 .423 .780 .367 .981 .078
Seat z .281 .258 1.000 .262 .294 .476 .451 .296 .801
Seat roll .466 .956 .262 1.000 .508 .844 .459 .978 .163
Seat pitch .971 .423 .294 .508 1.000 .667 .965 .483 .321
Seat yaw .608 .780 .476 .844 .667 1.000 .682 .820 .364
Back x .933 .367 .451 .459 .965 .682 1.000 .430 .451
Back y .450 .981 .296 .978 .483 .820 .430 1.000 .129
Back z .316 .078 .801 .163 .321 .364 .451 .129 1.000
Sig 1-tailed (p)

Seat x .000 .022 .066 .005 .000 .000 .000 .006 .045
Seat y .022 .000 .084 .000 .010 .000 .023 .000 .341
Seat z .066 .084 .000 .081 .057 .004 .006 .056 .000
Seat roll .005 .000 .081 .000 .002 .000 .005 .000 .195
Seat pitch .000 .010 .057 .002 .000 .000 .000 .003 .042
Seat yaw .000 .000 .004 .000 .000 .000 .000 .000 .024
Back x .000 .023 .006 .005 .000 .000 .000 .009 .006
Back y .006 .000 .056 .000 .003 .000 .009 .000 .248
Back z .007 .045 .341 .000 .195 .042 .024 .006 .000
184
The dependent variable is the averaged judgments of all subjects for each stimulus
(Table 42). The independent variables are the seat translational axes with frequency
weighting (without the multiplying factors).
Table 42. The selected dependent and independent variables used for calculating the
regression model. The frequency weighted r.m.s. values (without multiplying factors) for seat
translational axes are used as independent variables and averaged judgments of subjects for
each stimuli are used as dependent values.
Independent Dependent
Stimulus Seat x Seat y Seat z Judgment
1 0.308 0.284 0.702 0.709
2 0.192 0.162 0.701 0.509
3 0.154 0.107 0.704 0.445
4 0.429 0.401 0.706 0.998
5 0.404 0.215 0.297 0.441
6 0.149 0.108 0.273 0.216
7 0.250 0.803 0.349 0.974
8 0.163 0.705 0.158 0.845
9 0.082 0.659 0.133 0.639
10 0.270 0.554 0.765 0.924
11 0.153 0.272 0.718 0.464
12 0.103 0.199 0.728 0.409
13 0.741 0.803 1.197 1.622
14 0.400 0.417 1.126 1.113
15 0.302 0.238 1.135 0.887
16 0.335 0.389 0.458 0.697
17 0.171 0.208 0.432 0.341
18 0.123 0.130 0.437 0.342
19 0.472 0.435 0.331 0.891
20 0.225 0.323 0.297 0.440
21 0.111 0.171 0.257 0.212
22 0.473 0.246 0.240 0.778
23 0.351 0.163 0.287 0.467
24 0.341 0.087 0.277 0.499
25 0.620 0.915 1.461 1.415
26 0.349 0.461 1.371 1.082
27 0.211 0.210 1.389 0.941
28 0.296 0.147 0.317 0.483
29 0.660 0.216 0.243 1.267
30 0.352 0.550 0.286 0.733
185
8.4.1.3 Regression model 1
The results show that each independent variable had positive correlation to the
judgment (Table 43). The fore-and-aft axis had the strongest effect to the judgment, the
lateral the second and the vertical the lowest influence. Collinearity statistics show
good tolerance, so the correlation between independent variables did not affect the
results. P-value was below 0.05, so all results were significant.
Table 43. Multiple linear regression model parameters for model 1.
Unstandardised
Coefficients
Std.
B t Sig.
Error
Constant -.005 0.056 -.095 0.925
Seat x 1.118 0.152 7.378 <0.001
Seat y 0.640 0.108 5.907 <0.001
Seat z 0.275 0.061 4.510 <0.001
Table 44 shows the model summary. The correlation was calculated using Pearson’s
method.
Table 44. Summary for regression model 1 (Pearson).
2 Adjusted Std. Error of

r r 2
r the Estimate
0.942 0.888 0.875 0.124
The unstandardised coefficients (B) were used to form the regression model:
av = −0.005 + 1.118awx + 0.640awy + 0.275awz (27)
Where av is the discomfort value, awx is the frequency weighted r.m.s. value for the seat
As the coefficients are relative to each other, they can be scaled so that the seat
vertical coefficient is 1.0 (Table 45).
Table 45. Parameter transformation to comply with the standard multiplying factors, where
vertical axis has 1.0 factor.
Axis Original Scaled

Seat fore-and-aft (x) 0.275 4.1
Seat lateral (y) 0.640 2.3
Seat vertical (z) 1.118 1.0
186
Now the regression model can be described as:
av = −0.005 + 4.1awx + 2.3awy + 1.0awz (28)
8.4.1.4 Regression model 2
The results show that each independent variable had positive correlation to the
judgment (Table 46). The fore-and-aft axis had the strongest effect to the judgment, the
lateral the second and the vertical the lowest influence. Collinearity statistics show
good tolerance, so the correlation between independent variables was not affecting the
results. P-value was below 0.05, so all results were significant.
Table 46. Multiple linear regression model parameters for model 2.
Unstandardised
Coefficients
Std.
B t Sig.
Error
Constant 0.369 0.042 8.89 <0.0001
Seat x 1.426 0.242 5.89 <0.0001
Seat y 0.519 0.144 3.59 <0.0001
Seat z 0.184 0.048 3.79 <0.0001
Table 47 shows the model summary. The correlation was calculated using Pearson’s
method.
Table 47. Model summary.
2 Adjusted r Std. Error of

r r
Square the Estimate
0.917 .840 .820 0.1507
The unstandardised coefficients (B) were used to form the regression model:
av = 0.369 + 1.426awx
2
+ 0.519awy
2
+ 0.184awz
2
(29)
Where av is the discomfort value, awx is the frequency weighted r.m.s. value for the seat
As the coefficients are relative to each other, they can be scaled so that the seat
vertical coefficient is 1.0 (table 48).
187
Table 48. Parameter transformation to comply with the standard multiplying factors, where
vertical axis has 1.0 factor.
Axis Original Scaled

Seat fore-and-aft (x) 1.426 7.75
Seat lateral (y) 0.519 2.82
Seat vertical (z) 0.184 1.00
Now the regression model can be described as:
av = 0.369 + 7.75awx
2
+ 2.82awy
2
+ 1.0awz
2
(30)
The equation can be formulated to match the format of the brute force equation:
av = 0.369 + (2.78) 2 awx

2
+ (1.68) 2 awy
2
+ (1.0) 2 awz
2
(31)
Now the factors using regression model 2 are closely matched with the brute force
equation (model has no square root for the summed value, but it does not affect the
correlation as a ranking method is used).
8.4.2 Visual evaluation of clustering of multiplying factors
The results showed significant clustering of the seat horizontal multiplying factors using
the brute force method. This can be visualised using contour mapping (Figure 71). The
figure shows that the correlation (higher elevation) improved when the seat fore-and-aft
multiplying factor was between 2.0 and 3.1 and the seat lateral factor between 1.3 and
2.1. The figure also shows clear clustering of the best combinations and that the
standard multiplying factors (1.4) for the seat horizontal axes did improve correlation,
but not as much as higher factors.
188
Highest elevation
1.4 factors
Figure 71. Contour map of the seat translational axes multiplying factors where the vertical axis
is 1.0. The elevation (i.e. the correlation r) is increased for warmer colours.
The results show that the 1.4 factors did improve the correlation if compared to having
no factors, but increasing the factors improved the correlation even more (Figure 72).
Especially increasing the fore-and-aft factor the correlation improves significantly.
Fore-and-aft factor = 1.4

Lateral factor = 1.4
Figure 72. Left figure shows the contour map sliced from the seat lateral factor at 1.4 and 1.8.
Right figure shows curves sliced from the seat fore-and-aft factor at 1.4 and 2.7.
189
A more detailed contour map was created based on the previous contour map (Figure
73). The figure shows that the correlation was best (highest elevation) when the fore-
and-aft multiplying factor was between 2.6 and 2.9 and the lateral between 1.7 and 2.0.
Again clear clustering of the best values was evident.
Highest elevation
Figure 73. Contour map of the seat translational axes’ multiplying factors where the vertical axis
is 1.0.
Figure 74 shows the even more detailed contour map, where the highest elevation
shows that the correlation was best at fairly narrow set of multiplying factors where the
seat fore-and-aft multiplying factor was 2.6-2.8 and the lateral multiplying factor 1.7-
1.85. The results also explain why smaller step size produced better correlation, as the
highest peaks were between 0.1 decimal values.
190
Highest elevation
Figure 74. Contour map of seat translational axes’ multiplying factors where vertical axis is 1.0.
8.4.3 Improvement in correlation for new multiplying factors
The new models based on brute force method and linear regression models were used
to create Scenarios 8 (brute force), 9 (regression model 1) and 10 (regression model
2). Spearman correlation was calculated for all subjects and compared to the best
correlation found using the standard scenarios (Table 49).
The multiplying factors found using the brute force method (seat fore-and-aft 2.7,
lateral 1.8 and vertical 1.0) produced better correlation for 19 out of 22 subjects
compared to the best standard scenario. It also improved the average correlation for all
subjects to vibration, and the correlations of the "worst" and the "best" subject. There
was a clear improvement from the best standard scenario using Wilcoxon signed-rank
test (p<0.001).
The regression model 1 (Scenario 9) improved correlation for 11 out of 22 subjects.

The improvement was not statistically significant for the best standard scenario
(p=0.610). However, using regression model 2 (Scenario 10) the correlation was
improved for 17 out of 22 subjects and was statistically significant (p<0.005).
Comparison of the new scenarios showed that the multiplying factors for Scenario 8
produced the best results, although there was no statistical difference to Scenario 10
(p=0.078).
191
Table 49. Comparison between correlations of the best standard scenario and scenarios with
optimised multiplying factors for the seat translational axes (Scenario 8: seat fore-and-aft 2.7,
lateral 1.8 and vertical 1.0, Scenario 9: seat fore-and-aft 4.1, lateral 2.3 and vertical 1.0,
Scenario 10: seat fore-and-aft 2.78, lateral 1.68 and vertical 1.0). Scenarios 8 and 10 use root-
sum-of-squares method for combining the axes, while Scenario 9 uses direct summing.
Best standard scenarios Scenario 8 Scenario 9 Scenario 10

2 2 2 2
Test subject r Scenario r r r
1 0.420 Sc 6 0.471 0.446 0.469

2 0.611 Sc 6 0.610 0.568 0.598
3 0.612 Sc 6 0.615 0.574 0.602
4 0.635 Sc 7 0.650 0.646 0.663
5 0.694 Sc 6 0.752 0.722 0.735
6 0.585 Sc 6 0.560 0.530 0.561
7 0.468 Sc 7 0.484 0.504 0.484
8 0.710 Sc 6 0.803 0.752 0.804
9 0.396 Sc 6 0.409 0.368 0.394
10 0.518 Sc 3 0.531 0.480 0.544
11 0.552 Sc 7 0.584 0.564 0.579
12 0.586 Sc 7 0.710 0.702 0.706
13 0.645 Sc 6 0.714 0.636 0.706
14 0.731 Sc 7 0.751 0.730 0.748
15 0.642 Sc 7 0.674 0.641 0.678
16 0.635 Sc 6 0.680 0.688 0.676
17 0.653 Sc 6 0.716 0.722 0.713
18 0.597 Sc 7 0.567 0.551 0.557
19 0.664 Sc 6 0.707 0.686 0.693
20 0.496 Sc 7 0.548 0.513 0.545
21 0.614 Sc 7 0.621 0.609 0.617
22 0.485 Sc 7 0.493 0.484 0.512
8.4.4 Validating results
8.4.4.1 Correlation for two random groups
The validation was done by dividing the trial participants into two groups in random and
testing whether the results using the new multiplying factors (Scenario 8) will indicate
the same conclusions as for the whole group (Table 50).
192
Table 50. Participants divided into two groups.
Participant number
Group 1 Group 2
2 4
13 12
17 7
20 1
18 16
3 6
22 14
10 11
8 21
19 15
5 9
For the both random groups the multiplying factors found using the brute force method
were used as an additional Scenario 8 and compared against the best standard
scenarios (6 and 7) (Table 51). The results show that the new factors were better than
any standard scenario for the both groups and all together. Group 1 had systematically
higher correlation for all scenarios than group 2.
Table 51. Comparison between the best standard scenarios (6 and 7) and best new scenario (8)
for groups 1, 2 and all. Scenario 6 includes seat translational (with 1.4 multiplying factors for
fore-and-aft and lateral axes) and rotational axes (with respective multiplying factors), Scenario
7 includes seat translational and rotational axes and backrest translational axes with respective
multiplying factors, and Scenario 8 includes seat translational axes with 2.7, 1.8 and 1.0
multiplying factors (x, y and z).
2
Correlation (Spearman r )
Scenario Group 1 Group 2 All
Sc 6 0.873 0.820 0.844
Sc 7 0.891 0.810 0.850
Sc 8 0.971 0.902 0.950
8.4.4.2 Multiplying factors from other studies
Only one paper was found which suggested new multiplying factors for seat horizontal
axes (Maeda and Mansfield 2006b). Maeda and Mansfield proposed a 2.47 factor for
the both horizontal axes, which would match the discomfort of the vertical axis. The
results suggested that similar response was noticed for all axes, but because of
different weighting curves for the horizontal axes higher factors were needed to
compensate the difference.
193
8.4.4.3 Using new factors for data from other studies
The new set of multiplying factors (Scenario 8) for the seat translational axes were
tested using data from a previous trial from other researchers. Mansfield and Maeda
(2006) used stimuli which had equal energy from 1 to 20 Hz. The stimuli were made up
by having single-axis, dual-axis or tri-axial vibration. Three amplitude levels
(unweighted r.m.s.) 0.2 m/s2, 0.4 m/s2 and 0.8 m/s2 were used for generating 17
different stimuli. Each test subject was exposed to three repeats of the stimuli. The
subjects estimated the magnitude of each stimulus using Stevens’ method (i.e. the ME
method).
For comparison with the results in this study the unweighted values were converted to
weighted ones using 0.39 multiplier for the horizontal axes and 0.86 for the vertical axis
as described in the study (Table 52). Additionally the values were combined using
standard Scenario 2, where 1.4 multiplying factors were used for the seat horizontal
axes and Scenario 8, which include the new factors (fore-aft 2.7, lateral 1.8 and vertical
1.0).
194
Table 52. Summary of the weighted vibration magnitudes used by Mansfield and Maeda (2006)
(adapted from the source).
2
Vibration magnitude (m/s r.m.s.)
Stimulus x-axis y-axis z-axis r.s.s. scenario 1 r.s.s. scenario 2 r.s.s. scenario 8
1 0.078 - - 0.078 0.109 0.211
2 - 0.078 - 0.078 0.109 0.140
3 - - 0.172 0.172 0.172 0.172
4 0.156 - - 0.156 0.218 0.421
5 - 0.156 - 0.156 0.218 0.281
6 - - 0.344 0.344 0.344 0.344
7 0.312 - - 0.312 0.437 0.842
8 - 0.312 - 0.312 0.437 0.562
9 - - 0.688 0.688 0.688 0.688
10 - - 0.860 0.860 0.860 0.860
11 - - 1.032 1.032 1.032 1.032
12 0.156 0.156 - 0.221 0.309 0.506
13 0.156 - 0.344 0.378 0.407 0.544
14 - 0.156 0.344 0.378 0.407 0.444
15 0.078 0.078 0.172 0.204 0.231 0.306
16 0.156 0.156 0.344 0.409 0.462 0.612
17 0.312 0.312 0.688 0.817 0.925 1.224
Figure 75 shows correlation for all 15 subjects using scenarios 1, 2 and 8. In all cases
applying the multiplying factors (scenario 2 or 8) was better than not having any factors
(Wilcoxon p<0.05). For 9 out of 15 subjects scenario 8 was the best, but no statistical
differences were found to Scenario 2 (p=0.250).
195
2
Figure 75. Spearman correlation (r ) for each test subject for scenarios 1, 2 and 8.
8.5 Discussion
The results from the previous publications have indicated that the standard’s method
itself might not be valid. There have been doubts expressed at least with the frequency
weighting (Maeda, et al 2008), the multiplying factors (Maeda and Mansfield 2006b)
and the discomfort scaling (Maeda 2005). Griffin (1998) has criticised the current
standard of using the r.m.s. method instead of the r.m.q. or VDV methods. However, no
alternative methods have been proposed to replace the current standard regarding
discomfort.
Based on the analyses of the standard method in Chapter 7, it was evident that the
most effective way to improve the method is to optimise the multiplying factors. This
was also suggested by another publication (Maeda and Mansfield 2006b). The
analyses indicated that the seat horizontal axes should be given more emphasis and
that higher emphasis of the seat translational axes would minimise the need to use the
additional axes.
Although some publications have criticised the current standard, there have been no
suggestions on an improved model for predicting discomfort. The evidence in Chapter
7 showed that the frequency weighting did not have a significant role. This can be
because in practice several dynamic components act as natural low pass filter, thus the
stimuli only contains low frequencies. This might be one reason why laboratory studies
do not directly correlate to real environments.
196
Also other methods such as r.m.q. and VDV, which have been suggested to replace
the r.m.s. method when evaluating health effects, have not been shown to work better
for discomfort. Even though British Standard BS 6841 suggests using the r.m.q.
method for discomfort if crest factor of six is exceeded, it does not give any guideline or
values how to interpret the results.
8.5.1 Improvement in correlation using the new multiplying factors
The brute force calculations indicated that the seat translational axes alone can be
used to evaluate discomfort. This was expected as the backrest and rotational axes
had high correlation to the seat translational axes. Based on the calculations the seat
fore-and-aft axis should be emphasised 2.7 times, the lateral axis 1.8 times compared
to the vertical axis.
Even though the more detailed brute force search showed better results, the 0.01 step
size is considered an excessive level of precision, considering the limited size of the
data set used to determine the factors. The same practical accuracy was possible
using only the seat translational axes, thus the rest of the analyses was done using the
best factors derived from the first search. However the results showed that the seat
rotational and backrest axes do improve correlation with the proper factors, but the
biggest improvement of correlation is possible by optimising the multiplying factors for
the seat horizontal axes. The calculation time also dramatically shortens when using
lower step resolution.
The results showed that by increasing the multiplying factors of seat fore-and-aft and
lateral axes, the correlation improved systematically. The results showed similar
implications than Maeda and Mansfield (2006), which showed higher multiplying factors
(2.47) for seat horizontal axes than 1.4. The study concluded that frequency weightings
for seat horizontal axes underestimated the effect of vibration.
8.5.2 Confidence in the new multiplying factors
Both the brute force and linear regression model 2 showed similar results where the
fore-and-aft axis should be emphasised at 2.7 times and lateral axis 1.8 times
compared to the vertical axis when axes are summed using a second power method.
The brute force results showed clear clustering of the factors, which was evident from
the contour maps. This also explained why the same correlation was possible with
different combinations of factors. The results were also obtained using Kendall’s tau
197
correlation and similar results were found, thus the statistical analyses have a high
confidence level.
The validation of the results, by dividing the participants in random into two groups and
calculating the results again, showed that the new factors gave systematically better
correlation than the best standard scenario.
The results from another study were used to test the new factors. It was shown that the
higher factors improved correlation, but in this case there was no statistical
improvement compared to 1.4 multiplying factors (Scenario 2). The laboratory trial was
conducted using Steven’s magnitude estimation method and values were averaged
using geometric mean. These differences might have influenced the results. However,
the positive impact of higher multiplying factors was evident, and in average Scenario 8
was better than Scenario 2.
8.5.3 Comparison of results from different methods
Regression model 1 was based on linear summing, thus not directly comparable with
the brute force model. However, a similar trend in emphasising the fore-and-aft and
lateral axes was noted. Regression model 2 however had the second power summing,
thus the results were a close match to the brute force equation. As the square root
does not change the correlation using the Spearman method, the results indicated very
similar multiplying factors.
The both regression models gave results with a good confidence level. Even though
the regression model’s correlation was lower for model 2, it was using Pearson’s
correlation, which does not give best evaluation for this type of data. If Spearman
correlation was used, then model 1 had 0.894 (r2) and model 2 had 0.941 (r2). Thus
clearly the second power emphasis improved correlation to the discomfort. Thus model
2 was better and as the factors were so close with the brute force results, only one set
of multiplying factors (equation 31) was used for validating the results.
The regression model could not be used to include more than the seat translational
axes, as the method does not work when the independent variables have high
correlation. For the seat axes the regression model showed similar emphasis for the
axes. The factors of model 1 were higher than the brute force results, but this was
expected as the regression model equation sums the axes linearly without second
power emphasis. The higher factors for the both fore-and-aft and lateral axes
compensate this effect. Model 2 had practically the same multiplying factors than the
brute force model.
198
8.5.4 Applicability of the standard method
The whole purpose of evaluating discomfort from whole-body vibration is to understand

what characteristics in the vibration cause discomfort and how to minimise them. If the
method does not predict discomfort reliably it will not be used. Other methods such as
r.m.q. and VDV, which have been suggested as superior to the r.m.s. method when
evaluating health effects, were not shown to work better for discomfort for the range of
stimuli used in this study. It might be that different circumstances and environments will
lead to different emphasis of the axes, thus no generalised model can be realised. In
this case the model should allow changeable parameters for each type of environment
(e.g. train, car, boat). However, this study design used commonly experienced vibration
stimuli from road and off-road vehicles designed to capture the widest population of
those exposed.
Because the number of included axes will directly affect the size of OVTV, it is not
logical that axes can be neglected without compensation. Thus multiplying factors for
each combination of PVTVs should be introduced to allow comparison of results using
different number of axes. However, based on the results it can be speculated if
absolute discomfort level can be even predicted, as it will always depend on also other
factors than the vibration. It is more likely and practical that the method can be used to
predict relative discomfort between different vibration exposures. And it is likely that
causality between vibration and discomfort is not direct, thus only relative changes of
rather limited differences can be evaluated.
8.5.5 Scope and limitations of the new model
To avoid problems and misuse relating to the standard method, it is important to limit
the new method for a specific scope, instead of developing an “all purpose” model. This
improves the practical usability of the model. Based on the recommendations and
experience of the previous work the model should have the following constrains:
• The model only predicts relative change in discomfort relating to change in the
frequency weighted vibration amplitude for a seated person (with backrest);
o As other physical and physiological factors (i.e. noise and tiredness)

change absolute discomfort level, the method can only predict the
relative change related to the vibration if the other factors remain the
same.
• Following axes are needed to be measured (minimum requirement);
199
o A seat fore-and-aft, lateral and vertical axis;
• The vibration should not include high shocks (i.e. crest factor above 9.0);
It might be that different circumstances and environments will lead to different

emphasis of the axes, thus no generalised model can be realised. In this case the
model should allow changeable parameters for each environment. The best approach
is to use brute force, contour mapping and linear regression modelling to optimise the
factors.
8.6 Conclusions
It was previously concluded that the best approach to improve the current standardised
model would be to optimise the multiplying factors. The results here showed that:
• Higher multiplying factors for the seat horizontal axes improved the correlation
between vibration magnitude and discomfort;
• The seat translational axes alone could be used, if the multiplying factors were
optimised;
• The brute force and linear regression method showed similar results;
• Several combinations of the multiplying factors are possible as clustering of the

factors were noted;
• Most likely a good prediction is possible if relative change in discomfort is

evaluated instead of the absolute discomfort.
All analyses confirmed that the new multiplying factors gave better results than any
standard scenario. Based on the results of previous chapter a clear indication of
improvement has been found. The results should be validated in a field environment
using the similar method than in the laboratory trial.
200
9 Chapter 9 - Field trial: validating the
ISO 2631-1 method and results from
the laboratory trial
9.1 Introduction
It can be assumed that the ISO 2631-1 standard has been created for practical use.
Even though the method and procedure is based on laboratory studies, it should also
work in the field. However, it has not been widely adopted for predicting discomfort
from vibration. It is of importance to understand how well the method predicts
discomfort in field environments, where there are much larger possibilities for errors.
Also it is necessary to understand how to simplify the guidance to improve the
accuracy of the results and comparison to other measurements (the confusing
guidance has been discussed in Chapter 2).
It was previously noted in Chapter 2 that only a few 12-axis measurements have been
reported and only a few field trials have been made where discomfort to vibration has
been evaluated. Only one study was found where 12-axis vibration and subjective
judgments were simultaneously measured (Parsons and Griffin 1983). The study was
conducted using several cars on different road types. Subjects judged each leg lasting
20 seconds by marking a point in a line. The study was conducted on a closed test
track and each subject was tested separately. The method described in the study was
similar to that in the standard, but did not include multiplying factors and frequency
weightings were based on 8 subjects, thus not the same as in the standard. The results
from the study suggested that the r.m.s. method using root-sum-of-squares (r.s.s.)
summing is the best for car-type environments. The study weighted the axes based
only on the frequency weightings, thus no multiplying factors were used like in the
standards.
The results in previous Chapters 7 and 8 showed that it was possible to achieve good
correlation between the vibration exposure and subject’s judgment. Even though the
best standard scenario (all nine measured axes) already showed high correlation for all
subjects (r2 of 0.850), it was further improved (up to r2 of 0.950) and simplified by
introducing higher multiplying factors for seat horizontal axes (2.7 for the fore-and-aft,
1.8 for the lateral and 1.0 for the vertical axis). Using the aforementioned factors, only
the seat axes were necessary for achieving the best correlation. This was because of
high correlation between the seat and backrest translational axes due to the seat
201
structure and minimal effect of the seat rotational and floor translational axes to the
discomfort.
Based on the field measurements in Chapter 5, it was concluded that the seat
rotational and floor translational axes will have a marginal effect on the overall vibration
total value (OVTV) if standard weighting is used. Additionally, because of high
correlation between seat and backrest axes, it can be assumed that using the seat
translational axes will give as good results as using both the locations. This allows
simpler and more comparable results as a sensor location can vary significantly
between measurers and machines.
Field measurements will introduce several new factors and possibilities for errors,
which can be confined in a laboratory: stimuli cannot be completely controlled, noise
and other factors are present and will influence the judgment, repeatability depends on
the chosen tracks, which might change quickly, and so on. It might be that the vibration
is not the dominant factor for the discomfort in the chosen environment, thus results
might show low or inconsistent correlation to the vibration. There is not enough
information on multi-factorial effects on discomfort and how they relate. It might be that
separately evaluating each factor might give wrong indication of the discomfort in real
environments where all factors are present simultaneously.
9.2 Goals of the field trial

The purpose of the field trial is to validate if the standard method can be used in the
field for predicting discomfort from whole-body vibration using a similar approach to
that in the laboratory trial (Chapter 8). Answers to the following questions were
regarded as most important:
• Does the standard method adequately predict discomfort in the field?
o Adequately is defined as Spearman r2 > 0.5;
• Do the optimised multiplying factors found in the laboratory trial improve

correlation compared to the standard multiplying factors?
• What multiplying factors best describe the discomfort for whole-body vibration in
the field?
o Are the best found multiplying factors systematic within subjects and
groups?
202
9.3 Methods
9.3.1 Measurements
9.3.1.1 Field environment
The trial was conducted in Finland using normal road conditions around the Oulu
county area. The route consisted of asphalt, gravel, cobblestone and off-road surfaces
driven at various speeds (Table 53). The speed and location were recorded using a
GPS device. The legs were driven either in order from 1 to 11 or in reverse order from
11 to 1. The speed was maintained as accurately as possible whilst maintaining safety
on the public roads. The driver did not participate actively in data collection or
subjective rating, although vibration at their seat was measured.
Table 53. Legs used in the field trial.
Leg Surface Speed (km/h) Description
1 Asphalt 50 Three speed humps
2 Asphalt 45 Even city road
3 Cobblestonestone road 20 City centrum; pot holes
4 Cobblestonestone road 30 City centrum; pot holes
5 Asphalt 55 Even city road
6 Gravel 15 Rough gravel road with pot holes
7 Gravel 50 Gravel road with some pot holes
8 Off road 10 Uneven; pot holes
9 Asphalt 30 Speed humps
10 Gravel 20 Road construction site with loose gravel
11 Asphalt 110 Highway
Figure 76 shows the asphalt road type (legs 1, 2, 5, 9 and 11). Figure 77 shows the
cobblestone road surface (legs 3 and 4). Figure 78 shows off-road surface (leg 8).
Figure 79 shows gravel road (legs 6 and 7) and Figure 80 the construction gravel road
(leg 10). All road surfaces represented typical environments found in Finland. For
example a cobblestone road is typical in city centers in Finland and there are hundreds
of thousands of kilometres of gravel roads in the country side.
203
Figure 76. Asphalt surfaces used in the trial.
Figure 77. Cobblestone road surfaces used in the trial.
Figure 78. Off-road surface used in the trial.
204
Figure 79. Gravel road surfaces used in the trial.
Figure 80. Construction gravel road surface used in the trial.
9.3.1.2 Test subjects
Three test groups of seven subjects were used (Table 54). Groups 1 and 2 were driven
on the same day in October 2008, while group 3 was driven in June 2009. The route
was driven in same order for groups 1 and 3, while for group 2 the route was driven in
reversed order. Group 3 had four subjects from groups 1 and 2, and two new subjects.
The subjects were chosen and divided into the groups randomly. The average height
was 180 cm and weight 79 Kg, average age was 28 years. Most subjects were men.
Each test subject signed a consent form and was given information about the trial.
205
Table 54. The characteristics of the test subjects. Group 3 had subjects from groups 1 and 2
(same Subject id). Each seat had a fixed Seat id number.
Group Seat id Subject id Height (cm) Weight (kg) Gender
1 1 1 184 85 Male
1 2 2 182 90 Male
1 3 3 188 88 Male
1 4 4 190 100 Male
1 5 5 170 70 Male
1 6 6 191 85 Male
1 7 7 192 75 Male
2 1 9 190 78 Male
2 2 10 176 58 Female
2 3 11 170 63 Male
2 4 12 183 80 Male
2 5 13 184 95 Male
2 6 14 180 92 Male
2 7 15 178 72 Male
3 1 16 181 77 Male
3 2 17 155 60 Female
3 3 11 170 63 Male
3 4 5 170 70 Male
3 5 13 184 95 Male
3 6 14 180 92 Male
9.3.1.3 Measurement setup
Based on the goals of the trial, it was necessary to evaluate a number of subjects
simultaneously on different road surfaces. A car with several seats (i.e. minibus) was
considered a best option for conducting the measurements, thus a Volkswagen
Transporter was used (Figure 81). The vehicle had seven seats at the back and two
seats at the front. The back seats were used in the test. The test subjects sat on the
seats with a remote dial pad to indicate discomfort. The vibration was recorded
simultaneously with the discomfort judgment using the measurement device.
206
Figure 81. The test vehicle used in the field trial. Seven subjects was tested simultaneously.
Figure 82 shows the seating and measurement arrangement. The full 12-axis
measurements were conducted as a reference from the driver’s seat. From the
subjects the seat and backrest translational axes were measured along with the
discomfort judgment. Because the measurement device itself doubled as a backrest
sensor, the device could not be installed between the subject and the cushion. The
device was installed on the outer side of the backrest using an adhesive tape (Figure
83). Before the trial it was confirmed that the backrest sensor location gave practically
the same vibration levels as the correct location (i.e. sensor between a backrest and a
human) after frequency weighting.
4 5 6
X
2
1 3 7
Figure 82. Seating and measurement arrangement including seat ID numbers (top view).
Circles with diagonal patterns represent 6-axis measurement locations, a circle with a
horizontal pattern represents the 12-axis measurement location (driver).
207
Figure 83. Sensor configuration for each subject (left) and location of the backrest sensor (right).
9.3.1.4 Equipment
The trial was conducted using a commercial measurement device (TITAN by Vibsolas
Ltd) for gathering both the seat and backrest translational axes and the judgment data
from the test subjects, and a measurement device (HERCULES by Vibsolas Ltd) to
measure all twelve axes from the driver (Figure 84). TITAN saved acceleration data
from the axes and discomfort judgment at 256 Hz sampling rate. HERCULES saved
acceleration data from the twelve axes at 2000 Hz sampling frequency.
Figure 84. TITAN (left) and Hercules (right) measurement devices used in the field trial.
The continuous judgment method was previously chosen (Chapter 3) as the method for
acquiring judgment data, as it is feasible to use both in a laboratory and in a field for
long exposure durations. A development of the field version of the remote control dial
was required, which could be used with the TITAN device (Figure 85). The dial pad had
an indicator for the full range. The extremes of the knob where labelled as “no
discomfort” and “high discomfort”.
208
Figure 85. The developed remote dial pad given for each subject to judge discomfort. A tape
around the dial indicated scaling from “no discomfort” to “high discomfort”.
9.3.1.5 Trial procedure
Each group of seven subjects were driven through Oulu county area where there were
eleven legs each lasting between 90 and 120 seconds (Figure 86). The subjects
evaluated discomfort for each leg of the route separately. The subjects were instructed
to evaluate the discomfort continuously based on their most current feeling (i.e.
instantaneous judgment).
Figure 86. Oulu county area from up above and the whole test route (yellow line) used in the
trial.
Each leg was driven once. A measurement supervisor indicated the start and end of
each leg verbally. Between legs the subjects were instructed to relax and not talk to
each other. There was no communication between the subjects for the entire route.
The subjects did not see each other’s judgments. The windows were shaded with
blinds to decrease distraction. Also the test subjects did not see the driving direction as
209
an additional blind was installed between the driver and cabin. Figure 87 shows a test
subject conducting a judgment. Before the trial each test group was given a briefing of
the trial procedure in a meeting room. They signed a consent form and completed a
health screen questionnaire.
Figure 87. A test subject conducting a judgment during the measurements.
9.3.2 Analyses
Analyses of the data were conducted using similar principles as in the laboratory trial
(Chapter 6). Because no references were found using similar approaches for acquiring
judgment data and using longer exposure durations, it was necessary to analyse the
data from different perspectives.
9.3.2.1 Vibration magnitudes and frequency characteristics of the legs
To verify the stimuli types it was necessary to conduct analyses of the vibration
characteristics. Vibration magnitudes were calculated and evaluated based on the
standard method of the frequency weighted r.m.s. values without multiplying factors for
each axis. Also crest factors were calculated to verify the validity of results using the
r.m.s. method.
Frequency characteristics of the legs were analysed based on the power spectral
density figures. The frequency weighting curves were applied, as it was decided to
apply the standard weighting curves for the measured vibration for all analyses.
210
9.3.2.2 Effect of additional axes to overall vibration total value
Based on the assumption that the seat translational axes will effectively give the same
results using optimised multiplying factors, as using all twelve axes with respective
multiplying factors, it was necessary to validate how large a contribution the additional
axes have in this trial. Based on the previous results it was assumed that the seat
rotational and the floor translational axes have a relatively small contribution (less than
10 %). This was validated using 12-axis measurements from the driver in a similar
manner as in Chapter 5 using equations 21 and 22.
Even though the backrest translational axes have been noted to make a large
contribution to OVTV, they tend to have high correlation with the seat translational axes
(Spearman r2 > 0.90), thus the seat axes alone should give similar results with proper
multiplying factors. This was the result of the laboratory trial (Chapter 8). This was
validated in the field trial by calculating correlations for each axis from the seat and
backrest.
9.3.2.3 Analysis of judgment style
It was already noted in the laboratory trial (Chapter 6) that subjects had different styles
of judging: some considered their discomfort as cumulative from beginning of a
stimulus, while others judged instantaneous discomfort. In the laboratory trial there was
no effect of the judgment styles to the overall judgment value, as stimuli were short (15
seconds) and an averaged value of the last 10 seconds of each stimulus was
calculated. However, in the field trial stimuli were longer (up to 2 minutes) and several
judgment and vibration data pairs were calculated, thus the judgment style will be of
more importance. A visual analysis of different judgment styles was considered
necessary between the subjects and for the different legs. Based on the visual
inspection of the judgment styles, outliers were removed from the analysis.
9.3.2.4 Selecting best averaging window size and delay
The data were sampled at 256 Hz rate, thus the raw data needed to be averaged
before comparison of judgment data to vibration amplitudes was done. Previous
research using the continuous judgment method concluded an optimal averaging
window of 2.5 seconds with 1 second delay between vibration and judgment data
(Kuwano and Namba 1985). This conforms to the findings regarding the psychological
present, which a subject can consider as “one moment” without relying on a long term
211
memory. The optimal averaging size changes amongst subjects, thus different window
sizes should be considered to find a good compromise.
A pilot study noted a 2 second delay for reaction to a vibration stimulus (Bhalchandra
2008). This was confirmed in the laboratory trial (Chapter 6). This suggests trimming
each leg from the start. The references from noise studies and experience from the
laboratory trial can be considered as a basis for testing different combinations of
parameters for finding the optimal setup for analysing the correlation from the field
measurements.
Instead of using consecutive averaging windows, overlapping windows were

considered for improving the correlation. There are no references for using overlapping
windows with the continuous judgment method, thus the overlap was considered to be
half of the chosen averaging window size (Figure 88). The assumption was that the
overlapping improves the correlation and decreases the effects of different judgment
styles.
Figure 88. Illustration of overlapped values calculated from the vibration data.
Several setups were created for finding the best combination of parameters (Table 55).
The standard Scenario 2 (seat translational axes with 1.4 multiplying factors for the
seat fore-and-aft and lateral axes) was selected as it conforms best to the standard and
is the simplest form of the method. The parameters included testing different reaction
delays, averaging and overlapping window sizes. It was also considered necessary to
trim the data from the beginning of each leg similarly as in the laboratory trial (see
212
Chapter 6). The reason is to minimise differences between the subjects, because of
different delays at the start of each stimulus. A ten second trim was considered enough
based on a visual inspection to be sure that all subjects had reached their initial
discomfort level. Also a five second trim from the end of each leg was done to ensure
synchronised endings.
Table 55. Setups for testing best parameters for comparing correlation between vibration and
judgment values.
Setup Delay (s) Averaging size (s) Overlap size (s) Trim (start,end) (s)
1a 0 2.5 1.25 10,5
1b 2.5 2.5 1.25 10,5
2a 0 5.0 2.5 10,5
2b 2.5 5.0 2.5 10,5
3a 0 10 5.0 10,5
3b 2.5 10 5.0 10,5
4a 0 20 10 10,5
4b 2.5 20 10 10,5
5a 0 30 15 10,5
5b 2.5 30 15 10,5
As Spearman correlation was used, it was necessary to analyse how sample size
affects the correlation and what is the minimum sample size and correlation to be
accepted. This is the compromise of choosing the best setup from Table 55. Spearman
correlation gives more confidence if the number of samples is increased (Figure 89).
The figure can be used to determine the optimal number of sample sizes to minimise
the effect of outliers, but to still have good confidence in the results. It can be
concluded that at least a sample size of 22 should be considered as minimum. Similar
to the other chapters, Kendall’s tau was used to verify results of the Spearman
correlation. Wilcoxon signed-rank test was used to evaluate statistical differences
between the scenarios and the setups.
213
Figure 89. Spearman confidence table for determining the goodness of a correlation value
compared to a sample size.
9.3.2.5 Normalising the judgments
To calculate correlation between the vibration and the discomfort judgment for each
subject a normalisation was not needed as Spearman correlation is not affected by
absolute values (i.e. ranking method).
For averaging judgment data for each group a normalisation was required, because the
chosen method for judging vibration does not restrict a subject to evaluating discomfort
on a fixed scale. Normalisation of each test subject was done for the whole route by
considering the smallest judgment value as zero and highest as two, thus the judgment
values of each subject was converted to between 0.0 and 2.0. The averaging of the
judgment values for each group was done using an arithmetic mean.
9.3.2.6 Correlation between vibration magnitudes and discomfort
Correlation was calculated between the averaged judgment values and the frequency
weighted r.m.s. values of the applicable standard scenarios (i.e. scenarios using the
seat and backrest axes) (Table 56). Additionally the correlation was calculated for the
best set of multiplying factors found in the laboratory trial (Chapter 8). The correlation
analyses were made using Spearman r2.
214
Table 56. Practically realisable scenarios based on the ISO 2631-1 standard when using the
seat and backrest translational axes.
Scenario
Explanation
number
Overall vibration total value of seat translational axes (without 1.4 multiplying factors for
1
horizontal axes)
Overall vibration total value of seat translational axes (with 1.4 multiplying factors for
2
horizontal axes)
3
(without 1.4 multiplying factors) and backrest fore-and-aft axis (with 0.8 multiplying factor)
4
(without 1.4 multiplying factors) and backrest translational axes (with multiplying factors)
9.3.2.7 Optimising new multiplying factors
Using a similar approach to the laboratory trial, a best set of multiplying factors was
calculated for the data gathered in the field trial. This was accomplished using the brute
force method as it was concluded to produce the best results for finding the factors.
The multiplying factors were calculated for each subject and group. Only the seat
translational axes were used for simplifying the analyses. It was expected that
practically the same correlation can be achieved using the seat axes than the seat and
backrest axes. The correlation analyses were made using Spearman r2.
9.3.2.8 Clustering of multiplying factors
In the laboratory trial (Chapter 8) it was found that the brute force method produced
several sets of multiplying factors which gave practically the same correlation. This
suggested clustering of the factors. The contour maps were created, which confirmed
the findings visually. A similar approach was used also for the field data.
9.4 Results
All correlation analyses showed systematic results for both Spearman rho and
Kendall’s tau methods, thus results presented here are based only on the Spearman r2
values.
215
9.4.1 Vibration characteristics of the legs
9.4.1.1 Vibration magnitudes
Frequency weighted vibration magnitudes (r.m.s.) from the seat and backrest axes
showed moderate levels of vibration typically found in a car (Parsons and Griffin 1983)
(Table 57). The vibration magnitudes of most legs were less than 0.60 m/s2 except in
leg 8, which was off-road (dominant r.m.s. value was 0.75 m/s2). Thus the results of the
trial can be considered to represent passenger discomfort in a normal car and road
types. The vertical axis was dominant for all legs, which is normal when using a car.
Speed humps increased the vibration magnitudes significantly on asphalt surfaces.
There was a variation in the vibration magnitudes amongst subjects due to the seat
location, but the levels were consistent (i.e. high correlation).
Because of the seating arrangement the magnitudes were slightly different for each
subject. However, high correlation (Spearman r2 > 0.98) was found between the
seating locations. The vertical axis was highest for all legs, which is common in this
type of vehicle and selected road conditions. Also a standard deviation was highest for
the vertical axis. The seat and backrest axes had also high correlation (Spearman r2 >
0.90). Crest factors were below 9 for all cases.
2
Table 57. Frequency weighted r.m.s. values (m/s ) without the multiplying factors, and standard
deviations for all legs (without multiplying factors) averaged from subjects from group 1.
Seat Back
x y z x y z
Leg avg s.d. avg s.d. avg s.d. avg s.d. avg s.d. avg s.d.
1 0.298 ±0.106 0.130 ±0.025 0.533 ±0.155 0.574 ±0.141 0.158 ±0.023 0.619 ±0.130
2 0.104 ±0.024 0.073 ±0.010 0.197 ±0.050 0.164 ±0.037 0.095 ±0.012 0.305 ±0.042
3 0.250 ±0.070 0.256 ±0.052 0.503 ±0.129 0.450 ±0.117 0.300 ±0.035 0.574 ±0.094
4 0.261 ±0.068 0.196 ±0.044 0.541 ±0.159 0.481 ±0.118 0.265 ±0.049 0.793 ±0.117
5 0.086 ±0.026 0.085 ±0.011 0.228 ±0.063 0.164 ±0.048 0.116 ±0.024 0.399 ±0.050
6 0.295 ±0.077 0.383 ±0.070 0.513 ±0.156 0.498 ±0.159 0.432 ±0.033 0.562 ±0.130
7 0.146 ±0.033 0.249 ±0.036 0.588 ±0.195 0.274 ±0.056 0.365 ±0.083 1.231 ±0.167
8 0.349 ±0.083 0.624 ±0.109 0.748 ±0.145 0.633 ±0.146 0.702 ±0.048 0.759 ±0.124
9 0.288 ±0.068 0.171 ±0.023 0.470 ±0.117 0.522 ±0.116 0.199 ±0.019 0.554 ±0.100
10 0.234 ±0.048 0.301 ±0.053 0.564 ±0.143 0.425 ±0.105 0.372 ±0.032 0.891 ±0.108
11 0.171 ±0.040 0.119 ±0.017 0.407 ±0.114 0.319 ±0.092 0.173 ±0.035 0.631 ±0.095
216
9.4.1.2 Frequency characteristics
The legs used in the analysis can be divided in to five classes depending on the
surface type: 1) asphalt road with speed humps, 2) cobblestone road, 3) gravel road, 4)
off-road and 5) highway asphalt road. Each type was analysed using power spectral
density functions (Figures 90 - 92). Frequency weighted vibration data from subject 5
from group 1 was used for the analyses, as the location was in the middle of the
vehicle, thus best describes the overall vibration characteristics of the car.
The asphalt roads, with and without the speed humps, showed dominant frequencies in
1-2 Hz area (Figure 90). The highway asphalt showed resonance peaks over 10 Hz for
all axes, which is most likely due to the road characteristics. The gravel and
cobblestone roads showed similar characteristics for all axes: dominant frequencies
were below 2 Hz, but the lateral axis showed peaks also at 13 Hz. This was consistent
for both road types, and for the off-road as well. The frequency weighting curves
attenuated differences between the different legs, as they emphasise lower
frequencies.
Figure 90. Frequency weighted power spectral density function of asphalt road with speed
humps (legs 1 and 9) and highway asphalt (leg 11) for all translational axes from the seat
(subject 5 and group 1).
217
Figure 91. Frequency weighted power spectral density function of cobblestone road (legs 3 and
4) and gravel road (legs 6, 7 and 10) for all translational axes from the seat (subject 5 and group
1).
Figure 92. Frequency weighted power spectral density function of off-road (leg 8) for all
translational axes from the seat (subject 5 and group 1).
9.4.1.3 Effect of seat rotational and floor translational axes to overall vibration
total value
Twelve-axis measurements were made from the driver’s seat. The driver did not
participate in the judgment study, but served as a reference for measuring also the
floor and seat rotational axes. Although these were not used in further analysis they
confirmed that between 83 and 92% of OVTV occurred from the seat translational and
the backrest translational data, a result in agreement with the observations in Chapter
6. Therefore, the floor translational and the seat rotational data was not considered in
further analyses in this chapter, as they will have a marginal effect compared to the
complexity they cause for the measurements.
218
9.4.1.4 Correlation between the seat and backrest translational axes
Because of the measurement equipment used in the trial, the backrest sensor had to
be installed on the outer surface of the backrest. The sensor location was validated by
comparing the frequency spectra and r.m.s. values between the backrest sensor on the
driver (reference seat), which was installed correctly, and on the seat behind the driver
(seat id 6), where the sensor was on the outer surface of the backrest. Even though it
was already noted in Table 57 that the r.m.s. values varied amongst the seat locations,
the frequency spectra showed that similar trend was seen on both sensor locations
(Figure 93). This was validated for all legs and with and without frequency weighting.
Some differences were found depending on the leg and direction, but the dominant
frequencies had similar amplitudes.
Thus the backrest sensors of the subjects were used for analysis of the standard
scenarios. However, because of high correlation between the backrest and seat axes,
it was considered enough to use the seat translational axes for calculating the new
multiplying factors. This simplified the calculation process.
Figure 93. Frequency spectra (without weighting) for two different sensor locations for backrest
vertical axis. Blue line is for the reference location where the sensor was installed between the
subject and the cushion, and green line is where the sensor was installed outside the backrest,
behind the subject.
219
9.4.2 Characteristics of judgment response
9.4.2.1 Delay between vibration and judgment
Each leg lasted between 90 and 120 seconds and the test subjects were instructed to
estimate instantaneous discomfort using the remote dial. The dial was in “no
discomfort” position at the beginning of each leg, thus a delay was noted before the
subjects started to react to the vibration stimulus (Figure 94). In the laboratory trial this
delay was in average two seconds (and additional three seconds of adjustment period),
but in the field longer delays were noted. In some cases it took up to 15 seconds for the
subject to react to the current stimulus. The delay varied within the legs depending on
the road roughness and within each subject. The reaction to the vibration during the
legs was closer to 2 seconds, thus the delay of 2.5 seconds for the averaging was a
good estimate.
Delay Delay
Figure 94. An example of delay between the raw judgment data and the unweighted
acceleration for the vertical axis for subject 2 and for leg 3.
9.4.2.2 Analysis of judgment style
The study had several types of road surfaces. Some included shocks (i.e. speed
humps and pot holes). Others had more consistent vibration characteristics (i.e.
cobblestone and light gravel). It was concluded previously, that because of the different
vibration characteristics of the legs, the subjects might change their judgment styles
during the measurements. This will most likely affect the correlation when using a fixed
220
averaging size. Even though all subjects were instructed to judge based on their
instantaneous discomfort feeling, rather than overall discomfort from the start of a leg,
some subjects did judge the vibration more cumulatively than the others (Figure 95).
The figure shows differences of judgment styles for group 1, where subjects 5 and 7
judged more cumulatively than the rest (subject 6 was dismissed for technical problems
in the judgment data). However all subjects had a fast reaction to the speed humps.
Because of the smoothness of the road, the overall judgment level was low, but around
the speed humps high (i.e. speed humps dominated the discomfort judgment). This
was noticed for both legs with the speed humps (legs 1 and 9).
Figure 95. Judgment styles of group 1 for leg 1 with smooth asphalt and speed humps (subject
6 was not included, because of invalid data).
221
Legs 2 and 5 had smooth asphalt without the speed humps (Figure 96). Because of the
low vibration magnitudes throughout the legs the subjects did not show any meaningful
reaction to the vibration stimulus.
Figure 96. Different judgment styles of group 1 for leg 2 with smooth asphalt (subject 6 was not
included, because of invalid data).
222
In contrast leg 8 had high vibration magnitudes throughout the leg. In this case subjects
judged the leg more cumulatively (Figure 97). Again subject 5 had a different judgment
style than the rest, but in this case the difference was smaller compared to the other
subjects (i.e. the correlation was more similar).
Figure 97. Leg 8 judged by all subjects from the group 1 (subject 6 was not included, because
of invalid data).
Based on the legs it was clear that creating optimal windowed data pairs will require
different window and overlap sizes for different types of legs or the smallest usable
window size should be optimised based on the slowest judgment style. As a conclusion
it was noted that the shocks were judged instantly, but the slower the change in the
vibration also the more delay was in the judgment response.
223
The smooth asphalt roads, with and without the speed humps, showed different
judgment style than other legs (Table 58). In most cases each subject reacted similarly
for each leg, thus clear indication of systematic behaviour was found. The judgments
for legs 2 and 5 showed no proper judgments and the legs had too low vibration
magnitudes in contrast with the other legs.
Table 58. Summary of all legs and evaluated judgment styles.
Leg Description Judgment style

Three speed humps on a smooth Judgment style was primarily instantaneous as
1
asphalt background vibration was low
Judgment style was cumulative, but judgment level was
2 Smooth asphalt too low to use in the analysis. I.e. no real change from “no
vibration” case
Judgment style was cumulative with faster response to
3 Cobblestone road
high shocks
Judgment style was cumulative with faster response to
4 Cobblestone road
high shocks
Judgment style was cumulative, but judgment level was
5 Smooth asphalt too low to use in the analysis. I.e. no real change from “no
vibration” case
6 Gravel road with potholes Judgment style was cumulative
7 Gravel road with potholes Judgment style was cumulative
8 Off road Judgment style was cumulative
Judgment style was primarily instantaneous as
9 Smooth asphalt with speed humps
background vibration was low
10 Gravel Judgment style was cumulative
Highway asphalt at fast speed (110
11 Judgment style was cumulative
km/h)
9.4.3 Best setup and data
9.4.3.1 Best window and overlap size and delay for vibration and judgment
data
Because of a large number of different scenarios and setups, it was not feasible to
analyse and evaluate all combinations. Thus Scenario 2 (seat translational axes with
1.4 multiplying factors for horizontal axes) was chosen for evaluating which setup was
the best for the analysis. Scenario 2 was chosen as it had the best average correlation
of the standard scenarios.
Data from two subjects (3 and 14) from groups 1 and 2 are shown as an example of
visual evaluation (Figure 98 and Figure 99). Both subjects showed that clustering of the
data did not become significantly better after setup 4. Correlation of subject 3 did not
show any improvement using a delay, but subject 14 did have a small systematic
improvement with the delay. However, a visual evaluation showed that no practical
difference in clustering was noted with and without the delay for both subjects. Based
224
on the visual evaluation, it was concluded that setup 4b (window size 20 seconds,
overlap window 10 seconds, delay 2,5 seconds) would be the best option to conduct
analyses of correlation, as an inherent delay was noted also by the previous studies.
Even though setups 5a and 5b showed the best correlation values they had too few
samples for good confidence levels.
225
2
Figure 98. Correlation (Spearman r ) of different setups to a standard scenario 2 (the seat
translational axes with 1.4 multiplying factors for the horizontal axes) for subject 3. Setups are
explained in Table 56.
226
2
Figure 99. Correlation (Spearman r ) of different setups to a standard scenario 2 (the seat
translational axes with 1.4 multiplying factors for the horizontal axes) for subject 14 (group 2).
Setups are explained in Table 56.
227
Similar analysis was made for all subjects using the Spearman method (Table 59 and
Figure 100). The results show that the best setup (i.e. best correlation) varied for the
subjects. No systematic improvement was found using the delay, and it was noted that
although larger window sizes had higher correlation on average, not all subjects
showed improvement with a larger averaging window.
Analysing differences in more detail showed that in practice the correlation did not
systematically improve from setup 1a to 4b (W=-45, Z=-1.05 and p=0.294). A low
statistical difference was found between setup 1a and 5b (W=-80, Z=-1.88 and
p=0.060). Statistical differences between delays were noted only for setups 4 (W=-101,
Z=-2.38 and p=0.017) and 5 (W=-120, Z=-3.09 and p=0.002), which confirmed that
setup 4b would be the best compromise.
2
Table 59. Correlation (Spearman r ) for all setups for each subject using standard scenario 2
(the seat translational axes with 1.4 multiplying factors for the horizontal axes). Group 3 had
subjects also from the previous groups 1 and 2.
Setups
Subject Group 1a 1b 2a 2b 3a 3b 4a 4b 5a 5b
1 1 0.294 0.319 0.313 0.340 0.288 0.327 0.325 0.332 0.317 0.350
2 1 0.413 0.374 0.428 0.424 0.370 0.374 0.317 0.313 0.269 0.267
3 1 0.469 0.395 0.467 0.391 0.449 0.438 0.383 0.359 0.409 0.422
4 1 0.408 0.373 0.386 0.393 0.293 0.309 0.227 0.245 0.196 0.203
5 1 0.077 0.082 0.064 0.074 0.046 0.053 0.038 0.050 0.035 0.054
7 1 0.349 0.330 0.345 0.340 0.313 0.321 0.351 0.353 0.346 0.337
9 2 0.295 0.256 0.342 0.307 0.355 0.347 0.344 0.352 0.357 0.380
10 2 0.182 0.219 0.206 0.235 0.229 0.259 0.209 0.230 0.182 0.209
11 2 0.387 0.427 0.394 0.434 0.392 0.420 0.380 0.405 0.421 0.434
12 2 0.574 0.465 0.601 0.524 0.568 0.531 0.511 0.514 0.528 0.544
13 2 0.437 0.399 0.459 0.439 0.466 0.457 0.457 0.478 0.533 0.548
14 2 0.458 0.484 0.500 0.507 0.496 0.499 0.518 0.526 0.584 0.597
15 2 0.341 0.331 0.361 0.360 0.373 0.384 0.394 0.397 0.434 0.448
16 3 0.489 0.496 0.520 0.529 0.546 0.554 0.607 0.599 0.619 0.619
17 3 0.269 0.322 0.277 0.330 0.298 0.336 0.342 0.374 0.424 0.441
11 3 0.517 0.588 0.548 0.602 0.585 0.602 0.606 0.617 0.644 0.651
5 3 0.320 0.342 0.349 0.368 0.360 0.384 0.396 0.412 0.439 0.468
13 3 0.499 0.457 0.518 0.502 0.523 0.529 0.551 0.562 0.629 0.646
14 3 0.557 0.549 0.599 0.605 0.613 0.634 0.667 0.679 0.695 0.694
228
2
Figure 100. Correlation (Spearman r ) for all setups for each subjects using standard scenario 2
(seat translational axes with 1.4 multiplying factors for horizontal axes).
9.4.4 Valid subjects and legs for correlation analysis
Based on the evaluation of the correlation of the subjects and the previous evaluation
of the legs it was decided that some of the legs and subjects should not be used in the
analysis. The legs with mainly smooth asphalt at low speeds were discarded as the
subjects did not respond to the vibration level at all (legs 2 and 5). There was no data
from one subject due to a technical problem (6), and additionally subjects 5 (judgment
in group 1) and 10 (group 2) showed a judgment style, which were different from the
other subjects. This was confirmed when calculating the correlation (Table 59).
9.4.5 Normalising judgment data for comparison of subjects and groups
9.4.5.1 Each subject
The judgment data of the subjects were not normalised for calculating correlation for
individual subjects, as it did not have any effect on the results (Figure 101). The
averaging window size (20 seconds with 10 second overlap and 2.5 second delay)
attenuated single peaks, which improved correlation to the vibration (i.e. removed
outliers) and comparison between the different judgment styles.
229
Figure 101. The judgment data averaged using setup 4b (20 second window, 10 second overlap
and 2.5 second delay) for the whole route for subject 1.
9.4.5.2 Each group
For comparing the correlation of the groups a normalisation of the judgment data was
required for the subjects before averaging was possible (Figure 102). The
normalisation was done for each subject using the averaged data from setup 4a. Mean
values were calculated from the judgments of all subjects for each group.
Figure 102. Normalised judgment values of all subjects from group 1 (left) and an averaged
judgment value of the group (right) for setup 4b (20 second window, 10 second overlap and 2.5
second delay).
230
9.4.6 Correlation between vibration magnitudes and discomfort
9.4.6.1 Best standard scenario
The best standard scenario averaged from the correlations of all subjects, using setup
4b, was Scenario 2 (seat translational axes with 1.4 multiplying factors for horizontal
axes) although Scenario 1 had practically the same averaged correlation (p=0.682)
(Table 60). Even though Scenario 4 had lower averaged correlation, it was the highest
for six subjects from groups 1 and 2, and was not statistically different from Scenario 2
(p=0.294). Scenario 3 had lowest correlation from the standard scenarios (compared to
Scenario 2: p<0.001).
2
Table 60. Correlation (Spearman r ) for all scenarios for each subject. Group 3 had subjects
also from previous groups.
2
OVTV Scenarios (m/s )
Subject Group Sc 8 Sc 1 Sc 2 Sc 3 Sc 4
1 1 0.325 0.257 0.325 0.239 0.244

2 1 0.232 0.313 0.317 0.225 0.284
3 1 0.319 0.365 0.383 0.287 0.342
4 1 0.131 0.244 0.227 0.175 0.242
7 1 0.176 0.400 0.351 0.314 0.411
9 2 0.337 0.331 0.344 0.328 0.330
11 2 0.267 0.413 0.380 0.339 0.429
12 2 0.430 0.509 0.511 0.440 0.477
13 2 0.386 0.434 0.457 0.387 0.494
14 2 0.374 0.531 0.518 0.377 0.476
15 2 0.288 0.426 0.394 0.362 0.437
16 3 0.499 0.606 0.607 0.543 0.590
17 3 0.333 0.290 0.342 0.247 0.258
11 3 0.489 0.589 0.606 0.440 0.509
5 3 0.328 0.400 0.396 0.370 0.388
13 3 0.524 0.504 0.551 0.296 0.347
14 3 0.549 0.706 0.667 0.639 0.703
Average 0.324 0.400 0.401 0.329 0.388
231
Correlations for each group showed the same results, where Scenarios 1, 2 and 4 were
better than Scenario 3 (Table 61). It is notable that groups 1 and 2 had very similar
correlation, although they were driven in reversed order, and group 3 had
systematically better correlation. Group 3 did include four subjects from the previous
groups. The results were systematic for all groups.
2
Table 61. Correlation (Spearman r ) for all scenarios for each group
2
OVTV Scenarios (m/s )
Group Sc 8 Sc 1 Sc 2 Sc 3 Sc 4
1 0.386 0.554 0.553 0.464 0.534

2 0.398 0.522 0.504 0.451 0.551
3 0.652 0.788 0.766 0.678 0.756
Average 0.479 0.621 0.608 0.531 0.614
9.4.6.2 Optimised multiplying factors from the laboratory trial
For the field data it was noted that Scenario 8 (best multiplying factors found in the
laboratory trial) did not produce any better correlation than the standard scenarios
(Table 61 above). In fact the correlation was systematically worse than the best
standard Scenario 2 (p<0.001). Scenario 8 was also worse for each group (Table 61).
9.4.6.3 New multiplying factors
New multiplying factors were calculated for each subject separately (Table 62) and for
each group (Table 63) using setup 4b data. The multiplying factors were optimised
using the brute force method with the same principals as in Chapter 8.
The multiplying factors from the subjects showed clear consistency of the seat lateral
axis being the most dominant for the discomfort judgment. Even though Subject 17
showed different levels of values, the same trend was evident for all subjects. The seat
fore-and-aft had little or no correlation to the judgment compared to the lateral axis. All
subjects showed better correlation with the optimised factors than with any previous
scenarios. Compared with Scenario 2 the Wilcoxon test showed also statistical
difference (p<0.001).
232
Table 62. Optimised multiplying factors and correlation (Spearman) for each subject using the
brute force method. Subjects 5 and 6 from group 1, and 10 from group 2 were dismissed from
the analyses.
Multiplying factors for the

Frequency weighted r.m.s.
values for the seat
2
Subject Group r p x y z
1 1 0.659 <0.001 0.0 6.6 1.0
2 1 0.598 <0.001 0.0 11.4 1.0
3 1 0.593 <0.001 0.0 4.4 1.0
4 1 0.459 <0.001 0.0 8.0 1.0
7 1 0.501 <0.001 0.0 1.4 1.0
9 2 0.353 <0.001 0.8 1.8 1.0
11 2 0.545 <0.001 0.2 3.8 1.0
12 2 0.736 <0.001 0.0 3.4 1.0
13 2 0.600 <0.001 0.0 3.8 1.0
14 2 0.656 <0.001 0.0 2.2 1.0
15 2 0.514 <0.001 0.0 2.0 1.0
16 3 0.687 <0.001 0.0 2.8 1.0
17 3 0.611 <0.001 7.2 33.0 1.0
11 3 0.705 <0.001 0.0 2.2 1.0
5 3 0.476 <0.001 0.0 2.4 1.0
13 3 0.646 <0.001 0.0 2.8 1.0
14 3 0.761 <0.001 0.0 1.2 1.0
Average 0.553 <0.001 0.1 8.6 1.0
Median 0.600 <0.001 0.1 2.8 1.0
The optimised multiplying factors for each group were similar (Table 63). Again the
lateral axis was clearly dominant and the fore-and-aft axis had practically no correlation
to the judgments. Group 3 had the highest correlation, which can be linked to having
the subjects from the previous groups, which all improved the correlation from their first
run.
233
Table 63. Optimised multiplying factors and correlation (Spearman) for each group using the
brute force method.
Multiplying factors for the

frequency weighted r.m.s.
values for the seat
2
Group r p x Y z
1 0.686 <0.001 0.0 2.0 1.0
2 0.606 <0.001 0.2 2.0 1.0
3 0.846 <0.001 0.0 1.6 1.0
Average 0.713 <0.001 0.1 1.9 1.0
9.4.6.4 Clustering of the factors
A visual inspection of the correlation of the data points showed linear and positive trend
(Figure 103). It was evident that a correlation existed between vibration and judgments
for all groups, although Group 2 had somewhat worse agreement than groups 1 and 3.
Figure 103. Clustering of data points using new multiplying factors of each group using setup 4a
data.
Contour maps show that clustering of factors was similar for groups 1 and 3, while
group 2 had a wider range of combinations (Figure 104). All three maps show that a
small or zero value for the seat fore-and-aft axis gave the best correlation (i.e. the
lateral axis should have clearly larger multiplying factors than the fore-and-aft axis).
Highest elevation Highest elevation Highest elevation
Figure 104. Clustering of multiplying factors for seat horizontal axes (multiplying factor vertical
axis is 1.0) for the three groups using a contour map.
234
9.5 Discussion
The previous results indicated that the multiplying factors of the standard are not
optimal for evaluating discomfort from whole-body vibration. A new set of multiplying
factors were found to improve correlation in the laboratory trial (Chapter 8). Rarely
have results from a laboratory trial of this type been validated in a field. The purpose of
this field trial was to validate the standard and the results from the laboratory trial.
Similar to the laboratory study, calculation of new multiplying factors were considered
the best way to optimise the correlation and evaluate emphasis of axes.
The frequency weighting and r.m.s. averaging was chosen as the methods for
producing the vibration magnitudes, because they showed improved results in the
laboratory trial and they are required by the standard. Additionally the only reference
study conducted in a field (Parsons and Griffin 1983) concluded them to be the best
option as well.
9.5.1 Vibration magnitudes and frequency characteristics of the legs
The chosen legs showed typical vibration magnitudes of Finnish roads. The speed
humps and pot holes increased amplitudes in the asphalt and gravel roads, while the
cobblestone road showed high vibration magnitudes throughout the leg. Each leg was
driven in similar conditions for all groups, even though one group was driven 8 months
later than the others. No legs with only smooth asphalt were used in the analyses, as
they were noted to have too low threshold for any changes in discomfort.
Typical for a car environment, the vertical axis was dominant for all surface types. The
vibration magnitudes for the fore-and-aft and lateral axes were similar. The standard
deviation between the seat locations was small and systematic (i.e. higher magnitudes
showed higher standard deviation). Thus a very good correlation was found between
the seat locations. Asphalt roads without speeds humps will produce low vibration
magnitudes (less than 0.3 m/s2), but the speeds humps will clearly increase the r.m.s.
values up to 0.6 m/s2.
The legs were divided into five surface types for analysis of the vibration
characteristics. Because of the standard frequency weightings, there were no
significant differences in the frequency characteristics of the different surfaces. All legs
showed dominant frequencies below 2 Hz when the weighting was applied. This
implies that the role of the surface type (i.e. frequency characteristics) only have effects
below 10 Hz.
235
9.5.2 The effect of the additional axes
Based on the analysis the contribution of the seat rotational and floor translational axes
were considered small, thus they were not included in the evaluation of the correlation.
To simplify the measurements even more the effect of the backrest axes were
evaluated. Because of the high correlation to the seat axes, it was assumed that the
seat axes alone can be used to find multiplying factors for the optimum correlation. This
was based on the results in the laboratory trial (Chapter 8). This simplified the
measurements as the location of the sensors can cause errors, especially for the
additional axes. It is simpler to use the multiplying factors for the seat lateral axes to
compensate the effects of the additional axes.
9.5.3 Analysing different judgment styles and averaging the response
The method of continuous judgment has been used in noise research (Kuwano and
Namba 1985), which was used as a reference. However, in evaluating the annoyance
of noise only one parameter is considered, usually weighted A-value, thus the
correlation evaluation is more straight forward. Evaluating correlation to vibration is
more complex as several axes are included (three at minimum) and number of different
combinations are possible. This emphasises the need for simplifying the method so
that the analysis procedure is not too complex.
Based on the noise references, a 2.5 second averaging with 1 second delay was found
the best when analysing correlation. It was previously indicated that a reaction to
vibration will have at least 2 second delay (Chapter 6). Thus a 2.5 second reaction
delay was chosen to be tested. The results showed that the delay had marginal and
inconsistent effect for setups using a shorter window (less than 20 seconds). It
improved correlation for setups with 20 and 30 second windows (4 and 5). Setup 4b
(20 second window, 10 second overlap and 2.5 second delay) was used as it was a
good compromise of reasonable number of data samples and the correlation, and to
minimise differences in the judgment styles.
Response to vibration was noted slower when a leg did not show any sudden shocks.
The speed humps caused instantaneous change in the judgment, but in most cases
the response was between 2 and 10 seconds. This indicated that the data should be
calculated using a large enough window size, and overlapping windows. Detailed
analysis of each subject’s judgment data showed that some subjects had inconsistent
judgment style, thus influencing the results. This was noted for two subjects (5 and 10).
236
For each individual subject the judgment data was not normalised as a ranking method
was used for finding the correlation. However, when the correlation for each group was
calculated, a normalisation had to be done. There are no specific rules for this, but it
was reasonable to use each subject’s judgment from the whole route for normalising it
to between 0.0 and 2.0. Now all judgments from each subjects and group were able to
be averaged. There was a clear trend for all subjects, but it was still difficult to find a
good match for all subjects. However, as the correlation results from each group
showed similar trend compared to other groups and each subject separately, a
confidence in the method for using the judgment data was good.
It is clear that if using longer exposures in trials then guidance to the subjects on how
to evaluate discomfort is important. This requires more training than in laboratory trials.
Also it should be accepted that in a field the judgments and results in general will have
more errors, and it is important to analyse how to minimise the outliers so that they do
not skew the results, and thus conclusions. Averaging the continuous judgment data
correctly is an important factor.
9.5.4 Correlation between vibration magnitudes and discomfort
Several scenarios had to be tested as there were no references for continuous

judgements of discomfort and linking this to vibration. The scenarios used in the study
can be divided in to three categories: 1) scenarios based on the standard guidance, 2)
scenarios based on the optimised multiplying factors in the laboratory trial (Chapter 8)
and 3) scenarios based on the new optimised multiplying factors in this field trial.
It was found that using the standard guidance Scenario 2 produced best correlation,
but only Scenario 3, from the four standard scenarios, produced statistically worse
correlation. Scenario 8, which was based on the multiplying factors from the laboratory
trial showed similar correlation to Scenario 3, and was also statistically worse than
Scenarios 1, 2 and 4. However, the scenario using the new optimised multiplying
factors from the field data showed significantly improved correlation compared to any
other scenario. The results were systematic for individual subjects and for all groups.
A good correlation (r2 > 0.5) was found for the most subjects (14 out of 17) when using
the new multiplying factors. The correlation using the best standard scenario showed
good correlation for only 2 out of 17 subjects. Even though there was a variation
amongst subjects, the improvement in the correlation was systematic. A significant
improvement was noted also for all groups. The results from the groups showed the
same systematic behavior as the individual subjects.
237
Group 3 included subjects from both groups 1 and 2, and two new subjects. Comparing
correlation and judgment of the subjects, which had been exposed to the trial before,
even though 8 months apart, the correlation improved in all cases. This was noted for
all scenarios. It might be that exposing subjects to training beforehand will improve
their ability to judge the data. This has been noted in another study as well (Van der
Westhuizen and Van Niekerk 2006).
The reasons for the results of the different scenarios is based on one factor: the
multiplying factors of the fore-and-aft axes (the seat and the backrest). Where more
emphasis to the fore-and-aft axes was seen, there was worse correlation. This was
evident when the new multiplying factors from the field data was derived, which
showed zero or close to zero values for the fore-and-aft axes for the best correlation.
This was also the reason why the scenario using the multiplying factors from the
laboratory trial did not work, as it had high emphasis for the fore-and-aft axis (2.7). The
visual analyses of the clustering of the factors confirmed that the highest elevation for
the correlation was near zero for the fore-and-aft factor. This was similar for all groups.
The factors found in the laboratory trial suggested higher emphasis of the seat lateral
axis, but no emphasis on the fore-and-aft axis. Thus for some reason the subjects did
not respond systematically to the fore-and-aft motion. A simple reason for this was not
found, but the most likely cause for this is that in a car the fore-and-aft axis is more
present when accelerating and braking or going over the speed humps, where the
lateral and vertical axes are present more consistently.
9.5.5 Can discomfort be predicted practically using the standard or an

optimised method
Because of the implicit guidance of the standard a measurer would most likely use
Scenario 1 as a method for predicting discomfort. In this case none of the standard
scenarios would give good results, but Scenario 1 and 2 were the best options. The
reason was that because of the emphasis of the lateral axis compared to the fore-and-
aft axis, and both scenarios emphasise both axes the same. However, there was a
possibility to greatly increase the correlation if more adjustable method was used (i.e.
the optimisation of the multiplying factors).
The results were not systematic with the findings of the laboratory trial, which suggests
that the emphasis of the axes will change depending on the environment and the
vibration characteristics. For example in some environments there are no noticeable
fore-and-aft vibration, thus the correlation to the axis will most likely show inconsistent
238
results. However, the standard method implies that a fixed set of multiplying factors
should work in different environments. This might give a wrong conclusion of
discomfort in some environments. It is likely that a underlying factor has causality to
discomfort, which is not directly shown in the judgments. This limits the usability of the
method to assessing changes in environments which have similar characteristics (e.g.
noise and temperature).
Because there are strong indications that no fixed set of multiplying factors work for all
environments a guidance on how to measure and calculate new multiplying factors
should be given with the technical information. The method of continuous judgment
would be a good candidate as well as using the brute force technique as it is the most
robust method for finding the best combination of factors.
9.5.6 Recommended methods to evaluate discomfort from whole-body

vibration in field
Field environment generates many distractions (i.e. confounding factors), which are not
present in a laboratory. It is also harder to control the circumstances, as statistical
reliability is lower. Thus the method should be robust and simple enough to work also
in the field and give possibility for reasonable statistical reliability and comparability of
results.
9.5.6.1 What worked
The chosen method of continuous judgment was concluded to work well both in the
laboratory and in the field. It is designed to allow judgement of longer exposures than a
few seconds, and it captures the changes of judgment in great detail. It is important to
guide subjects for judging the vibration correctly. Based on the developed equipment it
was considered rather simple to acquire both vibration and judgment data from multiple
subjects simultaneously. The subjects were given short instructions and the route was
chosen from the local roads. The measurements took about an hour for each group.
Vibration magnitudes were relatively easy to measure from the seat and backrest
locations. Additionally it was noted, that also the 12-axis vibration can be measured in
a field with proper equipment. So it is possible to measure all axes in the field in
practice.
Exposure durations up to 2 minutes were accepted by the subjects and the

concentration was consistent throughout the measurement. However, not much longer
exposures are possible, if the subjects have to focus on the vibration constantly.
239
Spearman rho and Kendall’s tau are proven methods for assessing correlation where
normal distribution cannot be assumed. Both methods showed consistent results and
are easy to calculate with modern software programs.
9.5.6.2 What did not work
Some subjects will always not judge vibration “correctly”, thus will produce results that
are not in line with the other subjects. This requires manual inspection at this point
(until an algorithm is developed to automatically remove subjects). The method of
continuous judgment allows several judgment styles for the subjects, even though
guidance is given. This was evident in the judgment analyses (see Figure 95). It is
important to analyse how different judgment styles affect the correlation, as this helps
to improve the method in field.
Shorter gaps between the legs are recommended as some of the durations in this trial
were up to 10 minutes and it was possible that concentration levels dropped during the
final trials.
The route and legs should be considered carefully. In this trial some of the legs were in
a city center, thus traffic did influence the stimulus, thus creating differences between
the groups. Also exposure to low vibration magnitudes will show inconsistent
judgments, thus all surfaces should include at least medium vibration magnitudes (e.g.
from 0.3 to 0.6 m/s2).
9.5.7 Limitations of the results and model
The study covered only one car type with the same legs for each group. Different legs
or seats might produce differing results. This has been also noted previously (Griffin
2007). Also the driving style could cause differences, although a same driver was used
for all groups in the study. The stimuli were limited to the characteristics of normal
Finnish road conditions, thus not directly applicable to all work machine environments.
Also the seat was a conventional car seat with rigid body frame, thus not having any
suspension system (other than the seat cushion) between the seat and floor. Most
work machines have low frequency seat damping, which causes floor motion to be
significantly different than the seat or backrest motion.
Based on the earlier assumptions, only the seat and backrest translational axes were
measured from each subject, but additionally all twelve axes from the driver were
measured to verify the assumptions. The limitation of number of axes was also
necessary to allow simultaneous measurements of several subjects. Because of a
240
small number of data, it is still suggested that more 12-axis data is measured and
possibly similar measurements analyses in this trial would be repeated for using all
axes.
Even though the subjects from each group were exposed simultaneously, the vibration
magnitudes were different depending on the seat location. Closer to the axles (front
and back) the vibration was higher than in the middle of the car. This was a problem
when averaging the judgment values as also the vibration magnitudes from all subjects
needed to be averaged. As the correlation method (Spearman) is a ranking method
and because of the high correlation of the vibration between the seat locations, it was
possible to use the vibration data from one subject for comparison to an averaged
judgment value of a group.
The method of continuous judgment has no absolute scaling, thus each subject had
their own discomfort range. Because of relative long exposure durations (up to 2
minutes) and duration between legs (up to 10 minutes), the subjects might have
forgotten their scaling. However, this was not found to be significant, as correlations for
each subject was similar to that for the group.
The test subjects acted as passengers, thus the results cannot be directly used to
predict discomfort of the operator/driver. The operator’s discomfort can be affected also
by the complexity of the task and the work motivation (McLeod and Griffin 1993).
Because of the nature of the trial, noise and other factors could not be controlled. Thus
the judgments do include the effect of the other factors. However, as the test route and
the legs were same for all subjects, and the other factors remained somewhat
constant, the correlation between changes in the vibration characteristics and the
discomfort could be evaluated. There were some disturbances during the
measurements as some of the legs were in Oulu downtown and traffic changed the
experiment. However, as seven subjects were exposed to exactly the same
circumstances there was enough statistical data for the analysis.
9.5.8 Further work
The standard method has fixed multiplying factors for all axes without regarding
specific characteristics of the different environments. The factors define the emphasis
of each axis when calculating OVTV and should improve correlation compared not
using the multiplying factors. However, the results in this thesis has shown that
consistently multiplying the factors of the additional axes have degraded the
correlation.
241
It is not likely that a fixed set of multiplying factors would work best for all environments,
thus a subset of factors for specific environments should be calculated. The results in
this trial showed that the optimised multiplying factors could be used to predict
discomfort of subjects exposed to similar vibration characteristics. Further trials are
needed to evaluate how the optimimum multiplying factors change to conclude more
general trend.
This was a first study to acquire both vibration and judgment data in a field using the
continuous judgment, cross-modal matching method. Further trials need to be
conducted in different environments and surfaces to conclude, if multiplying factors will
change depending on the environment, or is there consistency in field. Also this type of
study should be conducted for measuring all twelve axes from each subject, although it
has been concluded that the effects of the additional axes can be overcome.
The results were not in line with the laboratory study in Chapter 8, where fore-and-aft
axis was concluded to be the most significant. The standard method can be used to
predict discomfort in a field if optimised multiplying factors are generated to each
specific environment. The multiplying factors in the standard have been derived based
on comfort contours found using single-axis sinusoidal exposures. Even though this is
the most fundamental approach, the results in this thesis have shown that in a multi-
axis environment the factors are not optimal. This might be because of several
reasons, thus a systematic research in a multi-axis environment using sinusoidal,
random and field-like stimuli is necessary.
9.6 Conclusions
A field study was made where both vibration and judgment data was gathered
simultaneously from groups of subjects. The conclusions were:
• Systematic trend of the correlation of the individual subjects and the groups
were found using the frequency weighted r.m.s. values;
• Significant improvement to the correlations using the standard guidance were

achieved using the multiplying factors optimised based on the field data;
o Higher emphasis to the lateral axis and lower emphasis to the fore-and-
aft was noted compared to the findings in the laboratory trial (Chapter
8);
• The results using the new multiplying factors improved correlation

systematically for all subjects and groups.
242
Based on the results of the laboratory and field trials several suggestions for improving
the standard method have been found. The complexity of the 12 axis measurements
have not been proven to give any practical improvement compared to using just seat
translational axes. The effects of the seat rotational and floor translational axes will be
small due to the weighting of the acceleration data and the axes. The high correlation
of the backrest axes to the seat axes in this type of seat structure will allow using just
the seat axes to evaluate discomfort.
243
10 Chapter 10: General discussion
10.1 Introduction
Exposure to vibration is an intrinsic part of our lives. The modern society is based on
moving around using cars, trains, ships, airplanes, etc. The last century has seen an
unprecedented growth in travelling. It is very hard to think of a situation or an
environment where there is no need to use any means of transport. There is almost no
place on earth, where a some form of a machine is not used for transporting people.
Practically all types of transport will cause vibration to a human body. In some cases
the vibration affects the human body so that permanent health effects are possible, but
luckily in most cases the vibration mainly causes discomfort. However, even low
vibration degrades work performance and hinders concentration. More and more
people also work while travelling, thus it is important to understand how to reduce
vibration in the most effective way.
The effects of vibration to humans can be assessed either subjectively or objectively.

Even though comments of a test subject give first-hand information on ride quality, it
requires resources and time to arrange tests and understanding of the factors (e.g.
physical size, background, motivation, etc) relating to the judgment. Thus, there is a
need for an objective method. An objective method can be used to predict the effects in
the field, a laboratory or a virtual world. A problem with an objective method is that it is
always limited by the methods, which have been used to develop it. It is thus important
to understand the background and assumptions of the objective method.
The international research community has acknowledged the importance in

understanding the effects of vibration by producing research results and several
standard versions. Many researchers use the standard methods or their own version
based on the standards to assess the effects on passengers and drivers in different
situations. However, since the early 1980’s no new method or any improvements to the
standard method have been proposed. For reasons unknown research on
understanding the effects of vibration on discomfort has been relatively small since
1986.
10.2 Benefits of understanding discomfort from vibration exposure

Discomfort is becoming more important to manufacturers of machines and vehicles, as
the acute health effects have been minimised. The work performance and comfort of
244
passengers can be improved, if the factors affecting the discomfort are known. There is
a growing number of people travelling for business and to work, and modern computers
have allowed working while travelling. Even small vibration magnitudes can hinder e.g.
reading and writing, thus reducing work performance. Currently, based on a perceived
lack of proven methods for assessing discomfort objectively, vehicle manufacturers rely
mostly on subjective questionnaires and methods, which have not been scientifically
proven (Kolich 2008). Thus the data has been difficult to compare and interpret. An
objective method will improve, systemise, accelerate and simplify the seat comfort
assessment through the development and manufacturing process.
10.3 Problems relating to the current ISO 2631-1 standard

The background behind the ISO 2631-1 (1997) standard method is from the studies
conducted in 1970’s and early 1980’s. The method as a whole was first introduced in
1986 in a SAE paper (Griffin 1986). It also became a British standard in 1987 (BS
6841). The method includes several parts, which translate vibration characteristics to a
value, which can be used to assess human reaction: 1) frequency weighting of
acceleration of each axis, 2) averaging the acceleration to a single value for each axis,
3) multiplying each axis by a factor and 4) combining the values of the axes to a single
value.
Before the BS 6841 standard, the previous standard methods were different and not
proven to be effective in practice (Wikstroem, et al 1991). There was very little field
validation on the methods. Even though the method in BS 6841 was shown to be
superior to the previous methods, it took ten years before a new version of ISO 2631
was agreed. It has been noted in publications that the reasons were more political than
methodical (Griffin 1998, Griffin 2007).
Even though the standard method is over 20 years old, there is very little
understanding of the practical accuracy of predicting discomfort from whole-body
vibration exposure. There are only few studies which have used the method to evaluate
discomfort and even fewer that have validated or tested different interpretations of the
method. Most validation and effort has been concentrated on the comfort contours,
which have been typically studied in a laboratory environment using discrete sinusoidal
stimuli, and typically in a single-axis. Other parts of the standard (i.e. multiplying factors
and averaging method) have had less validation. The few studies that were found
regarding the averaging methods showed inconclusive results, thus it is still not clear
whether the r.m.s. or VDV method should be prioritised. Also the studies have been
245
limited to specific environments, thus not allowing generalisation of the results. Only
one study was found to suggest higher multiplying factors for the seat horizontal axes
(Maeda and Mansfield 2006b), but no study was found, that had analysed the same for
all twelve axes.
The problems regarding the standard method relate more to the lack of information
rather than the presence of a better method. There is little information on how to select
the significant axes for each environment. The standard allows the use of a number of
different combinations, without giving any guidance on how to choose the proper one.
This may be one reason why so few studies have used all twelve axes. Another reason
is that measuring the twelve axes is technically so complex, that few have conducted
any in the field. Even if all axes were measured, the standard does not provide
practical information to assess the discomfort. Only a very crude table is given in the
standard, which is based only on limited data and has overlapping categories without
any detailed explanation.
Another interesting issue, which has not been discussed in other publications, is that
the size of OVTV is based on how many axes are included. Thus, OVTV will be smaller
if less axes are used, and based on the standard, the discomfort will be less. Of
course, the other axes do still exist and influence discomfort, but the method does not
have any way of compensating the missing axes from the assessment. An analysis in
Chapter 5 suggested compensating factors for each missing PVTV for producing OVTV
that would be comparable to OVTVs using different number of axes. It was found that a
highly linear correlation existed between PVTVs, thus a single compensating factor,
based on different numbers of axes, was produced. It is suggested that this type of
information is put in to the standard, so that the results would be more comparable.
This is not an issue when relative discomfort is assessed in the same environment, but
when absolute discomfort is compared to any kind of table with categories or to other
measurements.
10.4 Previous research and results

Most of the relevant studies were conducted between late 1960’s to early 1980’s. The
studies were mainly conducted in a laboratory using sinusoidal stimuli focusing on the
perception threshold and comfort contours. Most of the relevant studies were
conducted in the UK or Japan and by a handful of researchers. The findings led to few
publications, which defined and introducted the current standard method (Griffin, et al
246
1982, Griffin 1986, Parsons and Griffin 1983). Since then very little validation and
improvement of the method has been conducted.
Only few studies were found which have discussed or validated the standard method
for predicting discomfort from whole-body vibration. No studies were found that used a
multi-axis test bench linking judgments from subjects to OVTV. Additionally only a few
studies had actually measured and published results from 12-axis field measurements.
Only one study was found that measured twelve axes and discomfort judgment from
subjects in the field (Parsons and Griffin 1983). The results from that study showed
similar findings to the laboratory and field in this thesis: 1) although high individual
variability was found within subjects, a similar trend was noted, 2) there was little
difference between r.m.s. and r.m.q. methods, 3) more axes will improve correlation
and no single worst axis should be used, and 4) the method can be simplified, but
needs further understanding of how twelve axes are presented in field.
The literature review showed few studies which have tested the standard method in a
field environment using professional drivers (Fairley 1995, Wikstroem, et al 1991).
However, even in those cases there was a fixed test track, which emphasised shocks
more than those found in environments typical for passengers. The results might be
valid for large machinery, but cannot be directly used in other environments. There are
no studies where discomfort has been simultaneously acquired from several
passengers in an environment close to public transit.
Based on the literature review, it was concluded that more information is needed on
how the twelve axes are present in field environments, what are the effects of the
frequency weighting curves and multiplying factors, what is the practical correlation
between subjective judgments and vibration magnitudes using different numbers of
axes, is there a possibility to improve the standards, and do results in the laboratory
and in the field show similar findings.
10.5 Measuring vibration and subjective discomfort

An objective method does not require asking subjects to judge discomfort. However,
when evaluating the accuracy and statistical confidence of a method, then subjective
judgments should be acquired simultaneously with the vibration magnitudes. This is
more straight forward in a laboratory environment, but in the field several challenges
are faced. Both vibration and discomfort should be measured simultaneously as the
247
vibration exposure is constantly changing. Also it is much more difficult to observe the
subjects and control the environments, thus technical tools have more significance.
10.5.1 Measuring 12-axis vibration
The measurements are instructed to be conducted using acceleration sensors from a

seat surface, backrest and floor. Measuring all axes will require sensing both
translational and rotational accelerations. The translational axes are simple to measure
using 3-axis sensors, but the rotational axes need more complex configuration.
Because the standard method and guidance have not required the measurement of all
axes, there are very few commercial products available for measuring the rotational
axes.
The work in Chapter 4 showed that it is possible to produce affordable and simple
sensors for measuring all axes. There is no reason not to measure all axes as the
software for conducting the analyses is available and the effort to measure and analyse
all axes is not different from using just three axes from the seat surface. The work in
this thesis did produce a commercial 12-axis sensor system for conducting the health
and discomfort evaluation according to the ISO 2631-1 standard.
Based on the results in Chapter 5 it is highly recommended that more 12-axis data is
measured in the field from different machines and work phases. There is still very little
information on the contribution of the twelve axes, even though the results in Chapter 5
showed that it is most likely that the seat rotational and floor translational axes will have
insignificant effects independent of the machine type or an environment.
10.5.2 Acquiring judgment data
Different methods exist to acquire subjective judgements, while subjects are exposed
to vibration. However, there are only a few methods, which are suitable for a field
study. The methods discussed in the literature review in Chapter 2 were mainly
developed and are used in a laboratory, where a researcher can be present and take
notes. In the field it is hard to arrange for another person to take the notes, thus a
technical device should be used to allow independent recording of the judgments.
Normally the studies have exposed subjects to short stimuli (up to 10-15 seconds), and
one judgment for each stimulus has been recorded. This does not suit well for stimuli,
which are longer and represents the vibration found in the field better (i.e. non-
stationary and continuous). Only one method was found to suit the plans and goals for
this thesis. The continuous judgment method has been specifically developed for
248
acquiring judgments in a continuously changing environment. The method allows
researchers to analyse and calculate data, which is based on instantaneous and
continuous judgments. The device to allow digital recording of the judgment data
simultaneously to the vibration was easy to realise for both laboratory and field
environments.
The continuous judgement method has not been used in previous vibration research
but it has been validated in noise research. The results in Chapter 6 showed that the
method produced data which was systematic and did not require complex analysis.
The instructions for the subjects were short and an intuitive judging style was natural.
In the field trial (Chapter 9), more challenges for data processing was noted, which
related to the uncontrolled environment and long exposure time. However, it was
possible to optimise the results for analysing the judgment styles. The experience from
this thesis gave confidence in the method for future studies, especially relating to field
type environments and long exposures.
10.6 Relative contribution of the twelve axes in the field

The results from the field measurements in Chapter 5 showed, that because of
weighting for the rotational and floor axes, their contribution will be marginal in
practically all environments. This was evident in the laboratory trial (Chapter 7) as well,
even though the rotational axes were over emphasised. In practice, this means that
there is no need to measure the rotational and floor axes when calculating OVTV. A
small compensating factor (i.e. 1.1) can be used to convert OVTV using the seat and
backrest translational axes to match OVTV using all axes, as about 10% difference
was noted between the scenarios.
Chapter 5 combined results from two previous studies and the new field
measurements. The results included data from ten different machines, where some of
them were measured by all studies. Although there were large variations in vibration
magnitudes between the machines, they all showed similar trends in the dominant axes
and locations. Confidence in generalising the results was high, because all three
studies showed similar results, even though different equipment and environments
were used.
10.7 Using the standard method to predict discomfort

The literature review in Chapter 2 showed that little information exists on assessing the
confidence when using the standard method to predict discomfort. The several different
249
interpretations of the method, and the problems relating to the parts of the method,
have not given confidence when using it. Two trials were designed, in the laboratory
and field, for producing more information on the accuracy of the standard method.
10.7.1 laboratory trial
Based on the results in Chapters 7 and 8, it can be concluded that a high correlation
(Spearman r2 > 0.8) can be achieved between vibration stimuli and subjective
discomfort. Especially predicting the relative change in discomfort is possible, as the
results showed systematic behavior for all subjects. No differences were found
between men and women (i.e. physical size), thus a conclusion can be made that the
relative change in discomfort is more universal to all people than the absolute
discomfort.
The use of more axes was shown to improve correlation, but it was evident that the
seat translational axes alone will provide as good a correlation as all axes, if proper
multiplying factors are used. Even though the rotational axes improved correlation, their
effect was marginal, thus not justifying using them in most environments.
10.7.2 field trial
Based on the results in Chapter 9, it is possible to use the standard method with
optimised multiplying factors and understand discomfort in a field environment.
However, the results should be considered only in the environment where the study
has been taken, thus no generalisation is possible (i.e. multiplying factors cannot be
the same for all environments). There are so many variables in the field, that the
correlation will be weaker than in the laboratory. Still the correlation for each group of
people was found to be adequate (Spearman r2 > 0.5). The results showed that data
from one group could be used to predict discomfort of subjects in that environment, as
other groups showed a similar trend, even though the exposures were not in the same
order.
It was found that the axes from the seat and backrest correlated so well, that only the
seat translational axes needed to be used to predict discomfort. Similar to the
laboratory study, it was found that the rotational and floor axes had marginal effect to
OVTV, thus it was concluded that in most cases just the seat translational axes could
be used. This makes the field measurements more simple.
250
10.8 Improving the standard method
The results in Chapter 7 showed that depending on how the standard method was
applied, different correlations were produced between OVTV values and discomfort
judgments. Use of frequency weightings showed improvements in correlation, while the
multiplying factors degraded the correlation systematically. The r.m.s. method showed
better results than using the r.m.q. method. The correlation improved systematically for
all setups when more axes were included, however it was also noted that the
compensating factors did produce practically the same results as including the
compensated axes (i.e. using 1.4 multiplying factors to compensate the backrest axes).
10.8.1 Role of the axes
Only one study was found which measured all twelve axes and analysed the
correlation compared to different numbers of axes (Parsons and Griffin 1983).
However, in this study the only scenarios, that could be compared to the standard
method, were either using all axes, or only the worst axis. The results showed that
including all axes improved the correlation. The results in Chapter 7 showed similar
findings, where including more axes improved correlation for all different setups (i.e.
with and without the frequency weightings and the multiplying factors and different
averaging methods). It was also found that the axes from the seat and backrest had
high correlation, thus a compensating factors was calculated for each PVTV. On the
other hand, the role of the rotational and floor axes were found small and not providing
enough value compared to the effort of measuring them. Thus the conclusions from
Chapter 7 was that the seat translational axes alone could be used for predicting
discomfort, if the multiplying factors would be optimised. This was confirmed in
Chapters 8 and 9.
Can it be assumed that in every case all twelve axes will improve correlation, or are
there environments where some axes would actually degrade correlation with
discomfort? The method has its weaknesses, but there is no information, which
suggests that some axes will not work. The scenarios used in Chapter 7 showed that
the correlation was better in every case where more axes were added, but some axes
had more effect than others. The effect of the axes were best with the frequency
weighting, but without the multiplying factors, thus the role of the axes are not
independent from the frequency weightings and the multiplying factors.
251
10.8.2 Role of the frequency weighting curves
As most of the research and validation has been directed to the comfort contours (i.e.
frequency weightings), there was no evidence on changing them at this point. The
frequency weightings manipulate the frequency content of the measured data so, that
the effect to discomfort would be the most significant. The weightings for the most axes
attenuate frequencies above 4-5 Hz so much, that the frequencies above 10 Hz have
practically no effect. Only the weighting for the seat vertical axis and the floor axes (Wk)
include higher frequencies up to 20 Hz. It was noted in Chapters 5 and 9, that because
of the weighting curves, the frequency characteristics of the different environments
were practically similar. This will most likely result in very similar vibration magnitudes
as well.
The results in Chapter 7 showed, that very little difference was found in the correlation
with or without the frequency weighting the acceleration signals. In the field, the
frequency characteristics of a typical machine is similar to that of the frequency
weighting curves, thus no significant changes in the characteristics are seen. More
significant effects have been noted in laboratory studies, which can be linked to using a
sinusoidal or random flat-band stimuli, thus not representing the characteristics of a
vehicle in the field.
Based on the results it was concluded that the current frequency weightings have more
positive than negative effect to the correlation, thus the role of the multiplying factors
and averaging methods should be evaluated and optimised.
10.8.3 Role of the averaging method
As it is more feasible to assess effects of discomfort using a single value, a method to

average acceleration signals was introduced. The r.m.s. method, which is used in
many applications, was chosen early on as the main method for averaging the
acceleration signal. In some studies it was found that more emphasis on shocks should
be used, thus the r.m.q. method was proposed as an alternative. Both methods
average vibration over the time period, thus do not include information about the
duration. The VDV method was derived from the r.m.q. method to include also duration
as a factor.
Since creation of the ISO 2631-1 standard in 1997 no conclusive evidence supporting
either method was produced, so both methods were included. A crest factor method
was chosen to select which method is more appropriate, but even today there is no
252
concensus, of the best method. There is little information on the usability of the
methods relating to discomfort, especially outside a laboratory environment.
The results in Chapter 7 indicated no improvement in the correlation when emphasising

shocks. Even though the stimuli did not include extreme shocks, it had a wide range of
typical stimuli, which is found in the field. The few studies, which have suggested using
the r.m.q. (or VDV) method, have not analysed the correlation to discomfort, or have
purposely used stimuli with a high shock content. It has been also concluded that more
important than calculating the shock level is to understand the vibration in a more
statistical way.
10.8.4 Role of the multiplying factors
The purpose of the multiplying factors is to change the relative emphasis of each
frequency weighted vibration magnitude. The reason for introducing the factors were
the results from the comfort contour studies, where the same level of discomfort was
experienced for different levels of vibration magnitudes for different axes. To equalise
the effects, the multiplying factors were introduced, so that the axes could be combined
to a single value. However, the author was not able to find a publication where this
process was explained and the values for the factors derived. Thus the origin of the
multiplying factors is an interpretation of the author. The multiplying factors change the
emphasis of the different axes when calculating an OVTV. Each factor and vibration
value is taken to second power, thus the factors have great influence for PVTVs and
OVTV.
The results in Chapters 7, 8 and 9 showed that the role of the multiplying factors is
great when improving the standard method and understanding the relative contribution
of the axes. More emphasis on the horizontal axes from the seat were noted compared
to the vertical axis. Especially in the case where only the seat translational axes are
used, the horizontal factors should be higher than the suggested 1.4.
10.9 Implications of the findings for improving the standard method

This study used field-like stimuli both in the laboratory and in the field to validate and
optimise the standard method. This is a different approach than other studies, where
more controlled, simple and short stimuli were used. The results in Chapters 7, 8 and 9
suggested that the best approach for improving the correlation was to change the
253
Different sets of multiplying factors were derived from the laboratory and the field trials.
In both cases the emphasis on the lateral axes were found to be greater than in the
standard, but differences were noted for the fore-and-aft axis. In the laboratory a factor
of 2.7 was produced, where in the field practically zero value showed the best
correlation. The results implied two issues: 1) the multiplying factors for the axes will
not be the same for all environment and 2) in the field, the vibration to some axes is
either missing or produced from sources, which have inconsistent effects on discomfort
(e.g. speed humps, accelerating, turning, etc).
Even though there is validity in the fundamental approach of using sinusoidal stimuli in
a laboratory to minimise the other factors, the trials in this thesis showed that the
factors do not work in practice in a multi-axis environment and using field-like stimuli. It
is most likely that one general method does not work for all environments (actually it
has not been proven to be the best for any environment so far), so practical and
optimised versions of the method should be produced for different environments.
There is a difference in what type of behavior is behind the vibration magnitude value.
If the vibration is based on one high level movement (i.e. shock) or on more average
movement, it will change the judgment of the stimulus. The stimuli in the laboratory trial
(Chapters 6, 7 and 8) was generated to expose all to axes as many combinations as
possible. However, the stimuli in the field trial was not controlled in any way. The
results in the field trial showed, that there are environments, where one axis might have
different type of vibration characteristics than other axes, thus resulting in an
inconsistent correlation.
It might be that the comfort feeling from vibration exposure is so complex that it is not
possible to isolate the characteristics of it, or even isolate vibration from other factors
like noise. However difficult, there still needs to be an effort to further the knowledge,
because new information can significantly improve work environments. For doing this it
becomes critically important that the guidance for conducting the experiments is
simple, logical and universal, so that the information can be easily measured and
analysed, systematically compared and discussed.
254
10.10 An improved method for predicting discomfort from whole-
body vibration
Unfortunately this thesis did not conclude a single improved model for predicting
discomfort from whole-body vibration. However, it was concluded that a set of tools can
be used to optimise parts of the methods (i.e. the multiplying factors) for specific
environments and then using the derived factors to predict discomfort in that
environment. Thus the improved method includes steps first to produce the method
used for the prediction.
The following procedure is suggested for optimising the standard method to a specific
field environment:
1. A field trial is conducted for acquiring judgments and vibration simultaneously;
a. At least 21 subjects and selection of typical environments for producing

the vibration;
b. Exposures should be at least one minute long, and preferably over ten
different surfaces;
c. Three groups of seven subjects should be exposed to the stimuli in

different order;
d. Continuous judgment and vibration from the seat and backrest axes
should be measured simultaneously at 256 Hz sampling rate;
2. Analysing the correlation and finding the optimal multiplying factors;
a. The judgment data from each subject should be trimmed from the
beginning and end (5-10 seconds) and the rest of the data should be
averaged using different window sizes (10 to 20 second window size
with 2.5 second delay is a good start);
b. The averaged judgment data from all stimuli for each subject should be
normalised so that the minimum value is 0.0 and maximum is 2.0;
c. Vibration magnitudes should be calculated using the frequency

weightings and the r.m.s. method;
i. The same window sizes and trimming should be used to

calculate the r.m.s. values to pair the judgment and vibration
data;
255
d. Data for a group can be averaged using the normalised judgments from
the subjects and using vibration magnitudes from one subject;
e. The best possible correlation should be calculated using a brute force

method between all measured axes and the values for each subject and
group;
i. Number of different multiplying factors should be tested when

combining the vibration magnitudes of the axes to OVTV using
r.s.s. method;
f. The best correlation found for each subject and group should be
documented as well as a contour map produced on how the correlation
changes when the factors are changed;
i. Also the best correlations found using the standard multiplying

factors should be documented to assess the level of
improvement;
3. Assessing and using the improved method;
a. The new multiplying factors show the relative emphasis of the axes to
provide best correlation to discomfort;
b. The multiplying factors of the standard can now be replaced by the new
factors and objective measurement can be taken from similar types of
environments to predict discomfort.
10.11 Contribution of the thesis work

Even though the standard method has been available in practice since 1986, the full
method has been rarely used, thus little information is available on the practical use of
the method. The purpose of this thesis was not to develop a completely new method
for predicting the discomfort, but to understand, validate, improve and guide the use of
the standard method. The work in this thesis was intended to provide information from
all aspects of the discomfort assessment, from technical considerations for measuring
the vibration and judgment for optimising the method for a field use. Based on the
literature review, all parts required more information and analysis.
Since starting the thesis work in early 2006, few studies have been published on this
issue, but none of the studies have provided overlapping information or information
which would make parts of the results in this thesis obsolete. This thesis has provided
256
insight on the matter and can be used by other studies to compare the results and
understand how discomfort is changed under vibration exposure.
257
11 Chapter 11: Limitations and future
work
11.1 Limitations
Because of the chosen methods and hypotheses, and the scope of the trials, the
results in this thesis are limited. Also the method of the standard is limited, which was
the basis for this thesis. Despite the known problems relating to the standard method, it
was considered important to study and validate the method, and to find out how to
improve it. This is a valid goal as there are literally thousands of assessments done
using the method and all current commercial equipment is based on it.
Claims that a method has been validated are usually based on limited evidence of
consistency with some observation or impression in a specific situation and not
evidence that the method is appropriate over the full range of conditions to which it can
be applied (Griffin 2007). It is important also for the findings in this thesis to be limited
to the results and critically evaluate the usability of the results. The work in this thesis is
not related to a separate method, but is part of the research relating to validation of the
standard method.
11.1.1 Limitations of the standard method
It is possible that ISO 2631-1 does not provide a good model of subjective feeling
relating to vibration. Studies have found issues relating to frequency weightings
(Maeda and Mansfield 2006a, Maeda and Mansfield 2006b), multiplying factors
(Maeda and Mansfield 2006a), averaging methods and the guidance of the standard
(Griffin 1998). Griffin (1998) pointed out the problems and lack of references regarding
the use of Wk weighting and 1.4 multiplying factors for the horizontal axes. There
seems to be no valid studies behind them.
It has been shown that the current frequency weightings underestimate discomfort at
higher frequencies (above 30 Hz) and at higher magnitudes (i.e. larger errors at lower
magnitudes) (Griffin 2007). So the weightings are only reasonably valid in a specific
magnitude range, and they might be underestimating discomfort at magnitudes found
typically in a car or a train. The weightings are also derived from a large number of
results using statistical means, thus they are more representative of a population rather
than an individual (Mansfield 2005). It seems that weightings for the fore-and-aft and
lateral axes from seats underestimate the discomfort and should have higher
258
multiplying factors (Maeda and Mansfield 2006b). Also depending on the frequency
content of the vibration the discomfort scaling will most likely change (Maeda 2005).
This implies that no general absolute scaling can be provided, like it is provided in the
standard. The table provided in the standard (Annex C) provides an overlapping scale
of range of OVTV values that can be associated with a discomfort level (e.g. “high
discomfort”). The studies, which have been the basis for creating the table have all
used similar methods and have an inherent problem of producing overlapping
categories (Maeda 2005).
The multiplying factors derived for the twelve axes are mostly based on single-axis
experiments, where each axis and location has been separately analysed. There are
indications that a multi-axis environment will produce different results. This was also
shown by the laboratory and field trials in this thesis. However there are very few
results concerning multi-axis vibration and correlation to discomfort, especially using
field-like stimuli. Further experimental work is required in order to establish whether
there could be scenarios where it is necessary to measure rotational vibration.
Based on the earlier studies there is no consensus on the best averaging method. It is
interesting to note that the r.m.s. method seems to be better for low shock vibration, but
the r.m.q. and VDV for vibration including shocks. This limits the scope of the standard
method and separates it in to two versions.
11.1.2 Limitations of the theory behind the standard and applying the
standard in practice
The assumption is that discomfort relating to the vibration exposure can be evaluated
and predicted, and extracted from other factors like noise. This has to be assumed, as
otherwise there is no possibility to develop a method for evaluating the discomfort from
the vibration. However, it is not explicitly clear that this assumption is correct. It might
be that human decision making processing is too complex and non-linear to be
modeled using current approaches, especially in the field.
If assessment of discomfort also includes a behavioral component, such as changing to

different transport or vehicle, the calculated vibration values will not provide the
answer. This depends on the availability of the other transportation, costs and so on.
Thus the method can provide information on how to improve discomfort, but not to
assess the behavior of humans in a broader sense.
259
It was already speculated in the 1970’s that several receptors in a human body
perceive different characteristics of vibration, thus depending on the dominant
frequencies the receptors are different (Miwa and Yonekawa 1974). This might be one
reason for the vibration magnitude and frequency dependence of perception. In hand-
arm vibration research several receptors, which sense vibration at different frequencies
have been found (Mansfield 2005). In whole-body vibration a similar model of receptors
has not been suggested.
The applicability of the same method for different environments has been questioned
from the beginning of the first standard draft (Griffin 1978). It is more likely that the
discomfort is also linked inherently to other environmental variables than vibration, so
that just by examining vibration will produce differing results depending on the
environment. For example it is logical to assume that a paying customer (i.e.
passenger) will have different judgment criteria than a paid worker.
11.1.3 Limitations of the results
11.1.3.1 Laboratory experiments for validating the equipment
The equipment and sensor configuration for detecting rotational axes were validated in
the laboratory using a six-axis test bench and commercial equipment. A number of
different stimuli were used to verify the accuracy of the equipment at practical levels.
The experiments did not include a more detailed testing of the sensor dynamic
response at different frequencies, as it was assumed that the data provided by the
sensor manufacturer was correct. The same sensor was also tested in a different study
at the Technical Research Centre of Finland, but this was not done in a such detailed
manner, which is normally required for a measurement equipment.
11.1.3.2 Field measurements
Because of the limited time and resources only a number of work machines were
measured using all twelve axes. Additionally only a few repetitions were possible for
each work phase, because the measurements were made during the actual work. Thus
there was only limited data for concluding the results. This was noted also with the
previous publications having results of the twelve axes.
11.1.3.3 The laboratory trial
A laboratory study included 22 subjects and 30 different types of stimuli. Even though
the number of the subjects and stimuli were adequate for providing statistically
260
meaningful results, the test should be conducted in different environments and using
subjects from different backgrounds. The subjects in this trial were university students
and department staff, thus not accustomed to vibration exposure. Most of the subjects
were also young, which could affect the results.
The stimuli were different than normally used in laboratory experiments, as it was non-
stationary and based on field measurements. Thus the frequency characteristics of the
stimuli were dictated by realistic stimuli rather than theory. Each stimulus lasted 15
seconds, including continuously changing vibration characteristics. A similar kind of
study should be conducted using different equipment (i.e. test bench) and different
stimuli to conclude if the results can be generalised, or are specific to the environment.
Only one type of seat was used, which represented a typical car seat. The stimuli did
include vibration from work machines, but the seat might have changed the results so
that the results cannot be assumed to work as well in a work machine environment.
The subjects kept their feet on a non-moving platform, thus not all twelve axes were
used. Even though previous study using the same environment did not show any
differences when keeping the feet on the moving platform, the role of the feet should be
further studied.
The laboratory trial was limited to understanding the relative change in discomfort, thus
no absolute discomfort levels can be derived from the results.
11.1.3.4 The field trial
The standard method was tested using a car with subjects exposed to typical vibration
characteristics of Finnish roads. The trial was planned to validate the findings of the
laboratory trial. Several limitations had to be accepted to allow a trial in the field.
Only one car type was used in the trial. The car represented more of a taxi than a
normal car. The seats at the back where the subjects were seated, had less foam than
typical car seats. Thus the vibration was likely felt more than in a normal car. During
the measurements the weather was warm and sun was shining so that the
airconditioning was working at full. Depending on the seating location the effectiveness
of the air conditioning changed, thus some subjects might have been more warm than
others.
As already noted in the laboratory study, also subjects in this trial were young students
with few exceptions. The judgment styles showed that more detailed instructions
should be given and training of the subjects should be more emphasised.
261
11.2 Further work
The following list describes the suggested further work to validate and verify the results
in this thesis and to improve the understanding of the standard method:
• More 12-axis measurements in the field. It is necessary to validate and verify

if the effect of the seat rotational and floor translational axes is universal, and
whether there are some classes of vehicles where rotational motion could be
more significant (e.g. small boats). The measurements should be made using
all twelve axes and the frequency weighted vibration magnitudes using the
r.m.s. method should be published.
• Laboratory trials to verify the correlation between discomfort and

vibration magnitudes based on the standard method. Trials using different
type and duration of stimuli should be used to verify if similar results are found
than in the laboratory study in Chapter 7. Subjects with different background
and experience should be used to evaluate the characteristics of different
groups.
• Laboratory and field trials to verify the optimised multiplying factors and
the effects of the frequency weightings and the averaging methods. The
trials should be conducted using different types of stimuli from land and sea
vehicles to find the differences when optimising the multiplying factors. This
should lead to understanding similarities and differences between the
environments.
• Field trials to verify the need for different sets of multiplying factors. The
optimised multiplying factors derived from different environments should be
evaluated.
• Development of the method to acquire both vibration and judgment data

simultaneously and to calculate new optimised parameters for each
environment. Based on this thesis it seems most likely that the multiplying
factors (i.e. finding best possible correlation) will change for different
environments. Thus a simple method and technology should be provided for the
researchers to allow for the study of the environments.
• Revision of the standard guidance. The guidance needs to be simplified and

more information on assessing the results should be given. Also examples of
262
expected correlations should be presented as well as a general practical
guidance for the whole process.
263
12 Chapter 12 – Conclusions
12.1 Conclusions
This thesis has shown several important results, which will give a better insight for
using the standard ISO 2631-1 method to evaluate discomfort from whole-body
vibration in an optimal way. This thesis included two laboratory experiments to develop
and validate measurement equipment, a field study to measure all twelve axes from
several machines, and a laboratory and field trial to analyse discomfort from exposure
to multi-axis vibration. The following list shows the main conclusions derived from the
work (in the order of appearance in the thesis):
• There are very few studies and little information on how to use the
standard method to predict discomfort in practice. An extensive literature
review was conducted and showed the need for providing more information
about the standard method. The review found that there is no information on the
level of confidence expected for predicting discomfort in the laboratory or field.
Furthermore, several parts of the standard method and the theory behind it still
needs clarification and new information. Especially the information relating to
the applicability of the method in a field environment was missing.
• Affordable and simple equipment can be used to evaluate discomfort from

vibration using the full method. Several experiments and tests were
conducted for validating the sensor configuration for measuring rotational
accelerations. Also software was produced to allow simple evaluation of the
discomfort. The results showed that the full method can be used in practice and
does not require much more effort than measuring and analysing the three axes
from the seat.
• Seat and backrest translational axes are practically dominant in all field
environments when using the standard method. The most dominant axes
were seat vertical and backrest fore-and-aft. Additionally seat horizontal axes
showed high contribution, but less than the most dominant axes. The standard
method emphasises these axes so that it is likely that they are dominant in all
types of environments.
• Seat rotational and floor translational axes can be neglected for most
environments. The axes had only marginal (10 to 15 %) contribution to the
overall vibration total value when the standard frequency weighting and the
264
multiplying factors were used. As the contribution of the axes were within a
small range, the axes can be replaced by a compensating factor.
• The applicability of the r.m.s. method based on the crest factor is small if
all twelve axes are used. It was found that if all twelve axes are used, then it is
very likely that at least one axis will show a higher crest factor than the standard
limit for the r.m.s. method. For all twelve axes only a few measurements
showed crest factors below 9, but if the seat translational axes were used
alone, then only one-third of the measurements exceeded the limit. However,
the validity of the crest factor method can be suspected, especially for a longer
measurement duration, as it is based only on a single peak value.
• Different axes are dominant at different frequency ranges. It was found that
different axes are dominant depending on the frequency. Typically the seat and
backrest axes were dominant below and floor axes above 10 Hz. The dominant
axis changed multiple times from 0.5 to 80 Hz, thus the averaging of the time
signal might not provide enough information for accurate prediction of
discomfort.
• The effect of additional axes (i.e. point vibration total values) can be
compensated by using a single factor. Compensation factors were
calculated when less than twelve axes were included to calculate OVTV. The
results showed high collinearity between PVTVs, thus it was reasonable to
compensate missing PVTVs using a single factor. It is suggested to use a
compensating factor to allow direct comparison of overall vibration total values
with a different number of axes.
• The method of continuous judgment of line length can be used to

evaluate discomfort from whole-body vibration. The cross-modal method
was selected as a number of studies showed it to work well for non-stationary
long stimuli in noise research. It was also found to work well in vibration
research as the subjects easily adopted the technique to judge vibration, the
instantaneous judgment does not require long term memory, and the data can
be evaluated using different approaches. A delay was noted between
judgments and vibration magnitudes, which suggested a trimming of the data
and a small offset fix.
• There were no differences between men and women for judging relative
discomfort from non-stationary field-like stimuli. The results showed that
265
gender nor physical size were a factor when evaluating discomfort. Even
though previous publications have noted differences in gender, at least the
relative discomfort (i.e. change in discomfort when vibration characteristics
were changed) did not seem to be affected by gender.
• Correlation between vibration and discomfort improved when more axes

were used. Including all measured axes produced the best correlation for each
subject and for the whole group. All axes improved correlation, thus it is
important to include as many axes as possible. However, in most cases only
small improvement was noted compared to using seat translational axes, thus
greater complexity in analysis does not necessarily guarantee a better
correlation between the measured vibration magnitude and ratings of
discomfort.
• Frequency weighting curves improved correlation systematically. The use

of the frequency weighting was validated by comparing the frequency weighted
vibration magnitudes to the unweighted values. In each case, where more than
the seat translational axes were used, the weightings improved correlation.
• The r.m.s. method showed better correlation than the r.m.q. method. The
results showed no evidence that using the r.m.q. method would improve
correlation between vibration magnitude and discomfort. The r.m.s. method was
better systematically for all cases with and without the frequency weightings
and the multiplying factors.
• The multiplying factors degraded correlation for the additional axes. The
multiplying factors for the backrest and seat rotational axes degraded the
correlation for all setups. This was concluded based on the laboratory and field
trials. This was systematic for all subjects.
• Seat translational axes alone can be used to evaluate discomfort. It was

found that in both laboratory and field trials, practically the same correlation
between discomfort and vibration was achieved using only the seat translational
axes with multiplying factors. In the laboratory the 1.4 multiplying factors to
compensate the backrest axes provided practically the same correlation than
using all measured axes. The same scenario was also the best standard
scenario in the field trial. Thus it was concluded that the optimisation of the
multiplying factors should focus on the seat translational axes.
266
• Optimised multiplying factors improved the correlation to discomfort and
vibration exposure. The new optimised multiplying factors were calculated
using two different approaches (a brute force method and multiple linear
regression models) and the results were similar: the seat fore-and-axis should
be multiplied by 2.7 times and the seat lateral axis 1.8 times compared to the
vertical axis. The factors were higher than suggested by the standard when only
the seat translational axes are used (i.e. 1.4).
• The brute force technique and contours maps allowed easy optimisation
of the multiplying factors for different environments. The methods can be
used to calculate optimised multiplying factors for any environment where
judgment and vibration data have been previously recorded. The methods can
also be used to test how many axes are necessary to be measured.
• A developed device for simultaneously measuring vibration and

subjective judgments from multiple subjects was shown to work both in
laboratory and in field. Based on the requirements of the continuous judgment
method, a device was developed, which can be used to acquire instantaneous
judgments from subjects. The device can be used as an additional sensor, thus
synchronising the judgment with the vibration data. The subjects were able to
judge vibration effortlessly. This makes it possible to have stimuli, which is
longer (several minutes or hours) and less training of the subjects.
• A systematic trend in predicting correlation was found in the laboratory

and field. Even though the results from the laboratory and field trials showed
different sets of multiplying factors, the method itself provided similar trends for
each subject and group: the use of different multiplying factors improved
correlation.
• Significant improvement was achieved in correlation compared to the

standard method. In both laboratory and field trials a statistically significant
improvement was achieved using the optimised multiplying factors compared to
the best possible combination of axes based on the standard guidance. This
was evident for each subject and all groups.
• Discomfort can be predicted in the field using the standard method. The
predicted discomfort and the trend in optimising the correlation were the same
for different groups, even though the stimuli and route was not exactly the same
267
and one group was driven 8 months later. This shows high confidence that the
method can be used to predict relative discomfort in a specific environment.
• Discomfort should be evaluated only relatively within an environment. The

current standard implies that discomfort can be evaluated in an absolute scale
for a variety of environments. There is no evidence that supports this. The
results and conclusions in this thesis suggested that the current standard
guidance for evaluating discomfort should be changed to compare relative
discomfort within an environment.
268
13 References
Agius, R.M., Lloyd, M.H., Campbell, S.J., Graveling, R.A. and Hutchison, P.A. 1988,
Epidemiological and Clinical Studies of Back Pain in the Mining Industry, IOM
Research report, United Kingdom.
Ahn, S. and Griffin, M.J. 2008, "Effects of frequency, magnitude, damping, and
direction on the discomfort of vertical whole-body mechanical shocks", Journal of
Sound and Vibration, vol. 311, no. 1-2, pp. 485-497.
Alem, N.M., Hiltz, E.E., Breaux-Sims, A.M. and Bumgardner, B.A. 2004, "New
Methodology for Health Hazard Assessment of Repeated Shock in Military Tactical
Ground Vehicles", US Army research report, United States, pp. 9.
Beazant, G. 2005, "Europe wages war on bad vibes", Professional Engineering, vol.
18, no. 14, pp. 23.
Bhalchandra, J.A. 2008, Development of continuous subjective and physiological

assessment techniques for the determination of discomfort and activity-
interference in transportation system, Report, Loughborough University.
Boileau, P.-., Turcot, D. and Scory, H. 1989, "Evaluation of whole-body vibration

exposure using a fourth power method and comparison with ISO 2631", Journal of
Sound and Vibration, vol. 129, no. 1, pp. 143-154.
Boshuizen, H.C., Bongers, P.M. and Hulshof, C.T.J. 1990, "Long-term sick leave and
disability pensioning due to back disorders of tractor drivers exposed to whole-
body vibration", International archives of occupational and environmental health,
vol. 62, no. 2, pp. 117-122.
Bovenzi, M. and Betta, A. 1994, "Low-back disorders in agricultural tractor drivers

exposed to whole-body vibration and postural stress", Applied Ergonomics, vol.
25, no. 4, pp. 231-241.
Bovenzi, M. and Hulshof, C.T.J. 1998, "An updated review of epidemiologic studies on
the relationship between exposure to whole-body vibration and low back pain",
Journal of Sound and Vibration, vol. 215, no. 4, pp. 595-611.
269
Bovenzi, M., Pinto, I. and Stacchini, N. 2002, "Low back pain in port machinery
operators", Journal of Sound and Vibration, vol. 253, no. 1, pp. 3-20.
Bovenzi, M., Rui, F., Negro, C., D'Agostin, F., Angotzi, G., Bianchi, S., Bramanti, L.,
Festa, G., Gatti, S., Pinto, I., Rondina, L. and Stacchini, N. 2006, "An
epidemiological study of low back pain in professional drivers", Journal of Sound
and Vibration, vol. 298, no. 3, pp. 514-539.
Brereton, P. and Nelson, C. 2005, "The new vibration regulations and HSE's guidance
and expectations: whole-body vibration", 40th United Kingdom Conference on
Human Response to Vibration, Liverpool, England.
British Standards Institution 1987, British Standard Guide to Measurement and

evaluation od human exposure to whole-body mechanical vibration and repeated
shock, BS 6841:1987 edn.
Cation, S., Jack, R., Oliver, M., Dickey, J.P. and Lee-Shee, N.K. 2008, "Six degree of
freedom whole-body vibration during forestry skidder operations", International
Journal of Industrial Ergonomics, vol. 38, no. 9-10, pp. 739-757.
Cole, M.H. and Grimshaw, P.N. 2003, "Low back pain and lifting: a review of
epidemiology and aetiology", Work, a Journal of Prevention, Assessment and
Rehabilitation, vol. 21, no. 2, pp. 173-184.
Corbridge, C. and Griffin, M.J. 1986, "Vibration and comfort: Vertical and lateral motion
in the range 0.5 to 5.0 Hz", Ergonomics, vol. 29, no. 2, pp. 249-272.
Crichton, N.J. 1999, "Information point: Spearman's rank correlation", Journal of

Clinical Nursing, vol. 8, pp. 763.
Deprez, K., Moshou, D., Anthonis, J., De Baerdemaeker, J. and Ramon, H. 2005,
"Improvement of vibrational comfort on agricultural vehicles by passive and semi-
active cabin suspensions", Computers and Electronics in Agriculture, vol. 49, no.
3, pp. 431-40.
Donati, P., Grosjean, A., Mistrot, P. and Roure, L. 1983, "The subjective equivalence of
sinusoidal and random whole-body vibration in the sitting position (an experimental
270
study using the 'floating reference vibration' method)", Ergonomics, vol. 26, no. 3,
pp. 251-273.
Ebe, K. and Griffin, M.J. 2000a, "Qualitative models of seat discomfort including static
and dynamic factors", Ergonomics, vol. 43, no. 6, pp. 771-790.
Ebe, K. and Griffin, M.J. 2000b, "Quantitative prediction of overall seat discomfort",
Ergonomics, vol. 43, no. 6, pp. 791-806.
EC 2002, European Parliament Directive, 2002/44/EC on the introduction of measures

to encourage improvements in the safety and health of workers at work., Directive
edn, European Union, Brussels.
Els, P.S. 2005, "The applicability of ride comfort standards to off-road vehicles",
Journal of Terramechanics, vol. 42, no. 1, pp. 47-64.
EN 14253 2003, Mechanical vibration - Measurement and evaluation of occupational

exposure to whole-body vibration with reference to health - Practical guidance, .
European Comission 2007, Risks of occupational vibration exposures (VIBRISKS) final

technical report, EU.
Fairley, T.E. 1995, "Predicting the discomfort caused by tractor vibration", Ergonomics,
vol. 38, no. 10, pp. 2091-2091.
Fleming, D.B. and Griffin, M.J. 1975, "A study of the subjective equivalence of noise
and whole-body vibration", Journal of Sound and Vibration, vol. 42, no. 4, pp. 453-
461.
Fothergill, L.C. 1972, A study of the subjective response to whole-body vibratio, PhD
Thesis, University of Southampton.
Fothergill, L.C. and Griffin, M.J. 1977, "Evaluation of discomfort produced by multiple
frequency whole-body vibration", Ergonomics, vol. 20, no. 3, pp. 263-276.
Griefahn, B. and Broede, P. 1999, "The significance of lateral whole-body vibration

related to separately and sumltaneously applied vertical motions", Applied
Ergonomics, vol. 30, pp. 505.
271
Griffin, M.J. 2007, "Discomfort from feeling vehicle vibration", Vehicle System
Dynamics, vol. 45, no. 7-8, pp. 679-698.
Griffin, M.J. 1998, "A comparison of standardized methods for predicting the hazards of
whole-body vibration and repeated shocks", Journal of Sound and Vibration
International Conference on Whole-Body Vibration Injuries, 15-17 Sept.1997, vol.
215, no. 4, pp. 883-914.
Griffin, M.J. 1990, Handbook of human vibration, Academic Press, London.
Griffin, M.J. 1978, "Evaluation of vehicle vibration and seats", Applied Ergonomics, vol.
9, no. 1, pp. 15-21.
Griffin, M.J. 1976, "Subjective equivalence of sinusoidal and random whole-body

vibration", Journal of the Acoustical Society of America, vol. 60, no. 5, pp. 1140-
1145.
Griffin, M.J., Parsons, K.C. and Whitham, E.M. 1982, "Vibration and Comfort - 4.
Application and Experimental Results", Ergonomics, vol. 25, no. 8, pp. 721-739.
Griffin, M.J. and Whitham, E.M. 1980a, "Discomfort produced by impulsive whole-body
vibration", Journal of the Acoustical Society of America, vol. 68, no. 5, pp. 1277-
1284.
Griffin, M.J. and Whitham, E.M. 1980b, "Time dependency of whole-body vibration
discomfort", Journal of the Acoustical Society of America, vol. 68, no. 5, pp. 1522-
1523.
Griffin, M.J. and Whitham, E.M. 1977, "Assessing the discomfort of dual-axis whole-
body vibration", Journal of Sound and Vibration, vol. 54, no. 1, pp. 107-116.
Griffin, M.J. and Whitham, E.M. 1976, "Duration of whole-body vibration exposure: Its
effect on comfort", Journal of Sound and Vibration, vol. 48, no. 3, pp. 333-339.
Griffin, M.J., Whitham, E.M. and Parsons, K.C. 1982, "Vibration and Comfort - 1.
Translational Seat Vibration", Ergonomics, vol. 25, no. 7, pp. 603-630.
Griffin, M.J. 1986, "Evaluation of vibration with respect to human response", Passenger
Comfort, SAE, Detroit, MI, Engl, pp. 11.
272
Guilford, J.P. 1954, Psychometric methods, McGraw-Hill, New York.
Hacaambwa, T.M. and Giacomin, J. 2007, "Subjective response to seated fore-and-aft

direction whole-body vibration", International Journal of Industrial Ergonomics, vol.
37, no. 1, pp. 61-72.
Havelock, D., Kuwano, S. and Vorländer, M. (eds) 2008, Handbook of Signal

Processing in Acoustics, Springer.
Holmlund, P. and Lundström, R. 2001, "Mechanical impedance of the sitting human

body in single-axis compared to multi-axis whole-body vibration exposure", Clinical
Biomechanics, vol. 16, pp. 101.
Hostens, I. 2004, Analysis of seating during low frequency vibration exposure,

Katholieke Universiteit Leuven.
Hostens, I., Amditis, A., Stefani, O., Dangelmaier, M., Bekiaris, E., Schaerli, H.,
Bullinger, A. and Ramon, H. 2004, "SAFEGUARD seat/compartment evaluation
methodology for vehicles with suspended seats", Measurement Science and
Technology, vol. 15, no. 9, pp. 1742-1755.
Howarth, H.V.C. and Griffin, M.J. 1988, "The frequency dependence of subjective
reaction to vertical and horizontal whole-body vibration at low magnitudes",
Journal of the Acoustical Society of America, vol. 83, no. 4, pp. 1406-1413.
Hoy, J., Mubarak, N., Nelson, S., Sweerts De Landas, M., Magnusson, M., Okunribido,
O. and Pope, M. 2005, "Whole body vibration and posture as risk factors for low
back pain among forklift truck drivers", Journal of Sound and Vibration, vol. 284,
no. 3-5, pp. 933-946.
Hulshof, C.T.J., Verbeek, J.H.A.M., Van Der Laan, G. and Braam, I.T.J. 2002, "The fate
of Mrs Robinson: Criteria for recognition of whole-body vibration injury as an
occupational disease", Journal of Sound and Vibration, vol. 253, no. 1, pp. 185-
194.
Ishitake, T., Miyazaki, y., Noguchi, R., Ando, H. and Matoba, T. 2002, "Evaluation of
frequency weighting (ISO 2631-1) for acute effectsof whole-body vibration on
273
gastric motility", Second international conference on whole-body vibration
injuriesSiena, Italy.
ISO 2631/1 1985, Evaluation of human exposure to whole-body vibration: part 1 -

general requirements, International Organization for Standardization, Geneva.
ISO 10326 1992, Mechanical vibration: laboratory method for evaluating vehicle seat
vibration - part 1: basic requirements, International Organization for
Standardization, Geneva.
ISO 2631-1 1997, Mechanical vibration and shock: evaluation of human exposure to
whole-body vibration - part 1: general requirements, International Organization for
Standardization, Geneva.
ISO 2631-5 2004, Mechanical vibration and shock: evaluation of human exposure to
whole-body vibration - part 5: method for evaluation of vibration containing multiple
shocks, International Organization for Standardization, Geneva.
ISO 8041 2005, Human response to vibration -- Measuring instrumentation, 2nd edn,
International Organization for Standardization, Geneva.
ISO/TC 8/SC 4 2008, Mechanical vibration and shock - Evaluation of human exposure
to whole-body vibration - Part 1: General requirements - Amendment 1,
Amendment draft edn.
Jones, A.J. and Saunders, D.J. 1974, "A scale of human reaction to whole body,
vertical, sinusoidal vibration", Journal of Sound and Vibration, vol. 35, no. 4, pp.
503-20.
Jones, A.J. and Saunders, D.J. 1972, "Equal comfort contours for whole body vertical,
pulsed sinusoidal vibration", Journal of Sound and Vibration, vol. 23, no. 1, pp. 1-
14.
Kaneko, C., Hagiware, T. and Maeda, S. 2005, "Evaluation of Whole-Body Vibration by

the Category Judgment Method", Industrial health, vol. 43, no. 1, pp. 221-232.
Kjellberg, A. and Wikström, B.-. 1985, "Subjective reactions to whole-body vibration of

short duration", Journal of Sound and Vibration, vol. 99, no. 3, pp. 415-424.
274
Kolich, M. 2008, "A conceptual framework proposed to formalize the scientific
investigation of automobile seat comfort", Applied Ergonomics, vol. 39, no. 1, pp.
15-27.
Kumar, S. 1992, "Margin of safety for the human back: a probable consensus based on
published studies", Ergonomics, vol. 35, pp. 769-781.
Kuwano, S. and Namba, S. 1985, "Continuous judgment of level-fluctuating sounds

and the relationship between overall loudness and instantaneous loudness",
Psychological research, vol. 47, no. 1, pp. 27-37.
Kuwano, S. and Namba, S. 1980, "Continuous judgment of level-fluctuating sounds",

10th International Congress on Acoustics. Volume 3: Contributed Papers
Continued.Sydney, Aust, pp. B.
Lewis, C.H. and Griffin, M.J. 1998, "A comparison of evaluations and assessments
obtained using alternative standards for predicting the hazards of whole-body
vibration and repeated shocks", Journal of Sound and Vibration International
Conference on Whole-Body Vibration Injuries, 15-17 Sept.1997, vol. 215, no. 4,
pp. 915-926.
Lings, S. and Leboeuf-Yde, C. 1998, "Whole body vibration and low back pain
[Helkropsvibrationer og lænderygsmerter]", Ugeskrift for laeger, vol. 160, no. 29,
pp. 4298-4301.
Maeda, S., Kume, K. and Iwata, Y. 1983, "Evaluation of localized vibration using the
method of category judgment", Japan Industrial Management Assoc, vol. 34, no. 2,
pp. 200.
Maeda, S. and Mansfield, N.J. 2006a, "Effect of axis on subjective response of whole-
body vibration", The 9th western pacific acoustics conference, Seoul, Korea.
Maeda, S. and Mansfield, N.J. 2006b, "Multipliers of axes of whole-body multi-axis

vibration", 41st United Kingdom Group Meeting on Human Response to Vibration.
Maeda, S. 2007, The information on the measurements of 12-axis data by the author.
275
Maeda, S. 2005, "Necessary Research for Standardization of Subjective Scaling of
Whole-Body Vibration", Industrial health, , no. 43, pp. 390-401.
Maeda, S. 2004, "Comparison of 12 Axes Vibration Data on the Different Kinds of

Vehicle Seats According to the ISO 2631-1 Standard", 5th Int. Conf. on
Independent Component Analysis and Blind Signal Separation (ICA 2004).
Maeda, S., Mansfield, N.J. and Shibata, N. 2008, "Evaluation of subjective responses
to whole-body vibration exposure: Effect of frequency content", International
Journal of Industrial Ergonomics, vol. 38, no. 5-6, pp. 509-515.
Magnusson, M.L. and Pope, M.H. 1998, "A review of the biomechanics and
epidemiology of working postures (it isn't always vibration which is to blame!)",
Mansfield, N.J. 2005, Human response to vibration, CRC Press, London.
Mansfield, N.J. 2003, "Effect of increasing measurement duration on 'accuracy' of

whole-body vibration field measurements", 38th United Kingdom Conference on
Human Response to Vibration, Gosport, England.
Mansfield, N.J., Holmlund, P. and Lundstrom, R. 2001, "Apparent mass and absorbed
power during exposure to whole-body vibration and repeated shocks", Journal of
Mansfield, N.J., Holmlund, P. and Lundstrom, R. 2000, "Comparison of subjective

responses to vibration and shock with standard analysis methods and absorbed
power", Journal of Sound and Vibration, vol. 230, no. 3, pp. 477-491.
Mansfield, N.J., MacMull, S.J., Singla, G. and Rimell, A.N. 2007, "Relative influence of
sitting duration and vibration magnitude on sitting fiscomfort in a car seat", 42nd
United Kingdom Conference on Human Responses to Vibration, Southampton,
England.
Mansfield, N.J. and Maeda, S. 2005a, "Comparison of subjective ratings of whole-body

vibration for single and multi-axis vibration", 40th United Kingdom Conference on
Human Response to Vibration, Marriot Hotel, Queen Square Liverpool.
276
Mansfield, N.J. and Maeda, S. 2005b, "Equal sensation curves for whole-body vibration
expressed as a function of driving force", Journal of the Acoustical Society of
America, vol. 117, pp. 3853-3859.
McLeod, R.W. and Griffin, M.J. 1993, "Effects of duration and vibration on performance
of a continuous manual control task", Ergonomics, vol. 36, pp. 645-659.
Mehta, C.R. and Tewari, V.K. 2000, "Seating discomfort for tractor operators – a critical
review", International Journal of Industrial Ergonomics, vol. 25, no. 6, pp. 661-674.
Miwa, T. 1969, "Evaluation methods for vibration effect: part 8. The vibration greatness
of random waves", Industrial Health, vol. 6, pp. 11.
Miwa, T. 1967, "Evaluation methods for vibration effect. I. Measurements of threshold

and equal sensation contours of whole-body for vertical and horizontal vibration",
Industrial health, , no. 5, pp. 183-205.
Miwa, T. and Yonekawa, Y. 1974, "Evaluation methods for vibrations", Applied

Acoustics, vol. 7, no. 2, pp. 83-101.
Monsees, N., Whyte, R.T., Lines, J.A. and Stayner, R.M. 1988, "Relationship between
subjective assessment and objective measurements of tractor ride vibration (rms
and rmq)", Paper presented at the join French-British meeting on Human
Response to Vibration, INRS, Nancy, France.
Morioka, M. and Griffin, M.J. 2006, "Magnitude-dependence of equivalent comfort

contours for fore-and-aft, lateral and vertical whole-body vibration", Journal of
Namba, S. and Kuwano, S. 1988, "Measurement of habituation to noise using the

method of continuous judgment by category", Journal of Sound and Vibration, vol.
127, no. 3, pp. 507-511.
Namba, S. and Kuwano, S. 1982, "Psychological study on Leq as a measure of

loudness of various kinds of noises", Journal of the Acoustical Society of Japan,
vol. 38, pp. 774.
277
Namba, S. and Kuwano, S. 1980, "The relation between overall noisiness and
instantaneous judgment of noise and the effect of background noise level on
noisiness", Journal of the Acoustical Society of Japan, vol. 1, pp. 99.
Namba, S. and Kuwano, S. 1979, "Experimental study on the relation between long-
term annoyance and instantaneous judgment of level-fluctuating sounds",
Proceedings - International Conference on Noise Control Engineering, vol. 2, pp.
837-842.
Nawayseh, N. and Griffin, M.J. 2003, "Non-linear dual-axis biodynamic response to

vertical whole-body vibration", Journal of Sound and Vibration, vol. 268, pp. 503-
523.
Newell, G.S. and Mansfield, N.J. 2006, "Influence of Posture and Multi-Axis vibration
on Reaction Time Performance and Perceived Workload", 41st United Kingdom
Group Meeting on Human Responses to Vibration.
Newell, G.S., Mansfield, N.J. and Notini, L. 2006, "Inter-cycle variation in whole-body
vibration exposures of operators driving track-type loader machines", Journal of
Oborne, D.J. 1978a, "The stability of equal sensation contours for whole-body
vibration", Ergonomics, vol. 21, no. 8, pp. 651-658.
Oborne, D.J. 1978b, "Techniques available for the assessment of passenger comfort",
Applied Ergonomics, vol. 9, pp. 45.
Oborne, D.J. 1977, "Vibration and passenger comfort", Applied Ergonomics, vol. 8, no.
2, pp. 97-101.
Oborne, D.J. 1976, "A critical assessment of studies relating whole body vibration to
passenger comfort", Ergonomics, vol. 19, no. 6, pp. 751-774.
Oborne, D.J. and Clarke, M.J. 1974, "The determination of equal comfort zones for
whole-body vibration", Ergonomics, vol. 17, pp. 769-782.
Paddan, G.S. and Griffin, M.J. 2002a, "Effect of seating on exposures to whole-body
vibration in vehicles", Journal of Sound and Vibration, vol. 253, no. 1, pp. 215-241.
278
Paddan, G.S. and Griffin, M.J. 2002b, "Evaluation of whole-body vibration in vehicles",
Palmer, K.T., Griffin, M.J., Bendall, H., Pannett, B. and Coggon, D. 2000, "Prevalence
and pattern of occupational exposure to whole body vibration in Great Britain:
findings from a national survey", Occup Environ Med, vol. 57, no. 4, pp. 229-236.
Palmer, K.T., Griffin, M.J., Syddall, H.E., Pannett, B., Cooper, C. and Coggon, D. 2003,
"The relative importance of whole body vibration and occupational lifting as risk
factors for low-back pain", Occup Environ Med, vol. 60, no. 10, pp. 715-721.
Palmer, K.T., Harris, E.C., Griffin, M.J., Bennett, J., Reading, I., Sampson, M. and
Coggon, D. 2008, "Case-control study of low-back pain referred for magnetic
resonance imaging, with special focus on whole-body vibration", Scandinavian
Journal of Work, Environment and Health, vol. 34, no. 5, pp. 364-373.
Parsons, K.C. and Griffin, M.J. 1983, "Methods for Predicting Passenger Vibration
Discomfort", Passenger Car Meeting, SAE, Warrendale, Pa, USA, Dearborn, MI,
USA.
Parsons, K.C. and Griffin, M.J. 1982, "Vibration and Comfort - 2. Rotational Seat
Vibration", Ergonomics, vol. 25, no. 7, pp. 631-644.
Parsons, K.C. and Griffin, M.J. 1980, "Predicting the Vibration Discomfort of Seated
Passengers", Human Factors in Transport Research (Based on the Proceedings
of the International Conference on Ergonomics and Transport). Volume 2: User
Factors: Comfort, The Environment and Behaviour.Acad Press, London, Engl,
Swansea, Wales, pp. 114.
Parsons, K.C. and Griffin, M.J. 1978a, "Effect of rotational vibration in roll and pitch
axes on the discomfort of seated subjects", Ergonomics, vol. 21, no. 8, pp. 615-
625.
Parsons, K.C. and Griffin, M.J. 1978b, "The effect of the position of the axis of rotation
on the discomfort caused by whole-body roll and pitch vibrations of seated
persons", Journal of Sound and Vibration, vol. 58, no. 1, pp. 127-141.
279
Parsons, K.C., Griffin, M.J. and Whitham, E.M. 1982, "Vibration and Comfort - 3.
Translational Vibration of the Feet and Back", Ergonomics, vol. 25, no. 8, pp. 705-
719.
Parsons, K.C., Whitham, E.M. and Griffin, M.J. 1978, "Six axis vehicle vibration and its
effects on comfort", Transport Ergonomics, Conf, 1st, Feb 1978, vol. 22, pp. 211-
225.
Pelmear, P.L. and Leong, D.K.N. 2002, "EU directive on physical agents - vibration",
Journal of Low Frequency Noise, Vibration and Active Control, vol. 21, no. 3, pp.
131-139.
Pope, M., Magnusson, A., Lundstrom, R., Hulshof, C., Verbeek, J. and Bovenzi, M.
2002, "Guidelines for whole-body vibration health surveillance", Journal of Sound
and Vibration, vol. 253, no. 1, pp. 131-167.
Rehn, B., Bergdahl, I.A., Ahlgren, C., From, C., Jarvholm, B., Lundstrom, R., Nilsson,
T. and Sundelin, G. 2002, "Musculoskeletal symptoms among drivers of all-terrain
vehicles", 2nd International Conference on Whole-Body Vibration Injuries, Nov 7-9
2000, Academic Press, Siena, Italy, pp. 21.
Rehn, B., Lundstrom, R., Nilsson, T., Bergdahl, I.A., Ahlgren, C., From, C., Sundelin,
G. and Jarvholm, B. 2005a, "Symptoms of musculoskeletal disorders among
drivers of all-terrain vehicles in northern Sweden", Noise and Vibration Worldwide,
vol. 36, no. 1, pp. 13-18.
Rehn, B., Nilsson, T., Olofsson, B. and Lundstrom, R. 2005b, "Whole-body vibration
exposure and non-neutral neck postures during occupational use of all-terrain
vehicles", Annals of Occupational Hygiene, vol. 49, pp. 267-275.
Rehn, B., Nilsson, T., Olofsson, B. and Lundstrom, R. 2004, "Whole-body vibration
exposure and non-neutral neck postures during occupational use of all-terrain
vehicles", Annals of Occupational Hygiene, vol. 9, pp. 1-9.
Reiher, H. and Meister, F.J. 1932, "Sensitiveness of human body to vibrations", Power
Plant Engineering, vol. 36, no. 18, pp. 705.
280
Rybiak, C. 2003, The effects of posture, whole body vibration, and personal factors on
reported discomfort levels of operators during a simulated underground haulage
vehicle ride, Laurentian University of Sudbury (Canada).
Sandover, J. 1988, "Behaviour of the spine under shock and vibration: A review",
Clinical Biomechanics Behaviour of the Spine under Shock and Vibration, Jun 15
1988, vol. 3, pp. 249-256.
Sato, T., Maeda, S. and Yano, T. 2007, "Effects of noise exposure on whole-body
vibration perception threshold", Huanan Ligong Daxue Xuebao/Journal of South
China University of Technology (Natural Science), vol. 35, pp. 108-111.
Seidel, H. and Heide, R. 1986, "Long-term effects of whole-body vibration: a critical

survey of the literature", International Archives of Occupational and Environmental
Health (Historical Archive), vol. 58, no. 1, pp. 1-26.
Shoenberger, R.W. 1984, Subjective effects of combined-axis vibration: comparison of

Y-axis and Y-plus-roll vibrations, Aviat Space Environ Med 1984 May;55(5):387-
90.
Shoenberger, R.W. 1979, Psychophysical assessments of angular vibration:

comparison of vertical and roll vibrations, Aviat Space Environ Med 1979
Jul;50(7):688-91.
Shoenberger, R.W. and Harris, C.S. 1971, "Psychophysical assessment of whole-body

vibration", Human Factors, vol. 13, no. 1, pp. 41-50.
Sommerich, C.M., McGlothlin, J.D. and Marras, W.S. 1993, "Occupational risk factors
associated with soft tissue disorders of the shoulder: A review of recent
investigations in the literature", Ergonomics, vol. 36, pp. 697-717.
Speckhart, F.H. and Harrison, E. 1968, "Design of shock absorber to improve ride
comfort by reducing jerk", SAE MeetingSociety of Automotive Engineers (SAE), ,
pp. 7.
Stayner, R.M. 2001, Whole-body vibration and shock: A literature review, Technical
report, United Kingdom.
281
Stevens, S.S. 1957, "On the psychophysical law", Psychological review, vol. 64, no. 3,
pp. 153-181.
Sumitomo, S., Tsujimoto, S., Maeda, S. and Kitamura, Y. 1998, "The onfluence of the
great Hanshin earthquake on human response to environmental vibration due to
the shinkansen", Industrial Health, , no. 36, pp. 290.
Sundström, J. 2006, Difficulties to read and write under lateral vibration exposure, PhD
thesis, KTH, Sweden.
Teschke, K., Nicol, A.M. and Davies, S. 1999, Whole body vibration and back disorders
among motor vehicle drivers and heavy equipment operators. A review of scientific
evidence, Workers' Compensation Board of British Columbi, Vancouver, Canada.
Tyler, R. and Darlington, P. 2004, Measurement uncertainty in human exposure to

vibration, 39th UK Group Meeting on Human Responses to Vibration.
Van der Westhuizen, A. and Van Niekerk, J.L. 2006, "Verification of seat effective
amplitude transmissibility (SEAT) value as a reliable metric to predict dynamic seat
comfort", Journal of Sound and Vibration, vol. 295, no. 3-5, pp. 1060-1075.
Videman, T., Simonen, R., Usenius, J.-., Osterman, K. and Battie, M.C. 2000, "Long-
term effects of rally driving on spinal pathology", Clinical Biomechanics, vol. 15,
no. 2, pp. 83-86.
Wertheim, A.H. 1998, "Working in a moving environment", Ergonomics, vol. 41, no. 12,
pp. 1845-1858.
Whitham, E.M. and Griffin, M.J. 1977, Measuring vibration on soft seats, SAE paper,
US.
Wikstroem, B.-., Kjellberg, A. and Dallner, M. 1991, "Whole-body vibration. A

comparison of different methods for the evaluation of mechanical shocks",
International Journal of Industrial Ergonomics, vol. 7, no. 1, pp. 41-52.
Wilder, D., Magnusson, M.L., Fenwick, J. and Pope, M. 1994, "The effect of posture
and seat suspension design on discomfort and back muscle fatigue during
simulated truck driving", Applied Ergonomics, vol. 25, no. 2, pp. 66-76.
282
Wilder, D.G. and Pope, M.H. 1996, "Epidemiological and aetiological aspects of low
back pain in vibration environments - an update", Clinical Biomechanics, vol. 11,
pp. 61-73.
Wyllie, I.H. and Griffin, M.J. 2007, "Discomfort from sinusoidal oscillation in the roll and
lateral axes at frequencies between 0.2 and 1.6 Hz", Journal of the Acoustical
Society of America, vol. 121, no. 5, pp. 2644-2654.
Yamashita, K. and Maeda, S. 2003, "Realization of 12-Axis Vibration Measurement on

the Seat according to the ISO 2631-1 Standard", The 32nd International Congress
and Exposition on Noise Control Engineering, Jeju International Convention
Center, Seogwipo, Korea.
283

Validation and Improvement of The ISO 2631-1 (1997) Standard Method For Evaluating Discomfort From Whole-Body Vibration in A Multi-Axis Environment

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Validation and Improvement of The ISO 2631-1 (1997) Standard Method For Evaluating Discomfort From Whole-Body Vibration in A Multi-Axis Environment

Uploaded by

Copyright:

Available Formats

Loughborough University

Validation and improvement

• This thesis is not yet available. A Doctoral Thesis. Submitted in partial

Metadata Record: https://dspace.lboro.ac.uk/2134/6250

Publisher: c Yka Marjanen

Please cite the published version.

For the full text of this licence, please go to:

Submitted in partial fulfilment of the requirements for the award of

© by Ykä Marjanen (2010)

Keywords: Discomfort, whole-body vibration, standard, ISO 2631-1, multi-axis,

I have been fortunate to have resources to participate in many conferences concerning

My parents and grand parents deserve also mentioning, as I am who I am because of

TABLE OF CONTENTS ................................................................................... VI

LIST OF TABLES ........................................................................................... XVI

LIST OF FIGURES ......................................................................................... XXI

1 CHAPTER 1 – INTRODUCTION .................................................................. 1

1.1 BACKGROUND .................................................................................................. 1

2 CHAPTER 2 - LITERATURE REVIEW ........................................................ 9

2.1 INTRODUCTION AND AIMS OF THE REVIEW .......................................................... 9

3 CHAPTER 3 – PLANNING AND SELECTING GENERAL METHODS FOR

3.1 INTRODUCTION ............................................................................................... 72

4 CHAPTER 4 - DEVELOPING AND VALIDATING EQUIPMENT FOR

4.1 INTRODUCTION ............................................................................................... 77

5 CHAPTER 5 - FIELD MEASUREMENTS: CONTRIBUTION OF TWELVE

5.1 INTRODUCTION ............................................................................................. 100

6 CHAPTER 6 - LABORATORY TRIAL PART I: DESCRIPTION AND

6.1 INTRODUCTION ............................................................................................. 137

7 CHAPTER 7 - LABORATORY TRIAL PART II: VALIDATING THE ISO

7.1 INTRODUCTION ............................................................................................. 160

8 CHAPTER 8 - LABORATORY TRIAL PART III: IMPROVING THE ISO

8.1 INTRODUCTION ............................................................................................. 178

9 CHAPTER 9 - FIELD TRIAL: VALIDATING THE ISO 2631-1 METHOD

9.1 INTRODUCTION ............................................................................................. 201

10 CHAPTER 10: GENERAL DISCUSSION ............................................. 244

10.1 INTRODUCTION ............................................................................................. 244

10.11 CONTRIBUTION OF THE THESIS WORK ......................................................... 256

11 CHAPTER 11: LIMITATIONS AND FUTURE WORK .......................... 258

11.1 LIMITATIONS ................................................................................................. 258

12 CHAPTER 12 – CONCLUSIONS ......................................................... 264

12.1 CONCLUSIONS ............................................................................................. 264

TABLE 3. SUMMARY OF FIELD STUDIES REGARDING SUGGESTED METHODS FOR

TABLE 6. APPROXIMATE INDICATIONS OF LIKELY REACTIONS TO VARIOUS MAGNITUDES

TABLE 10. THE THEORETICAL SENSITIVITY (I.E. SMALLEST DETECTABLE CHANGE IN

TABLE 24. THE RELATIVE CONTRIBUTION (%) OF FREQUENCY WEIGHTED SEAT

TABLE 27. COMPARISON OF OVERALL VIBRATION TOTAL VALUES (M/S2) OF SEAT

TABLE 28. COMPENSATING FACTORS FOR DIFFERENT COMBINATIONS OF POINT VIBRATION

TABLE 35. PRACTICALLY REALISABLE SCENARIOS FOR CALCULATING AN OVERALL VIBRATION

TABLE 36. THE CORRELATION (R2) BETWEEN DISCOMFORT JUDGMENTS OF INDIVIDUAL

TABLE 50. PARTICIPANTS DIVIDED INTO TWO GROUPS. .................................................. 193

FIGURE 3. MEDIAN EQUIVALENT COMFORT CONTOURS OF DIFFERENT STUDIES FOR VERTICAL

FIGURE 7. MIWA’S STUDIES ON SEAT TRANSLATIONAL AXES USING VIBRATION GREATNESS

FIGURE 10. COMPARISON OF BS 6841 STANDARD WEIGHTINGS TO NEWER FINDINGS.

FIGURE 15. COMMERCIAL 12-AXIS MEASUREMENT EQUIPMENT FOR DISCOMFORT

FIGURE 16. 12-AXIS MEASUREMENT SYSTEM INSTALLED ON A BUS SEAT ACCORDING TO

FIGURE 21. MEASUREMENT AND ANALYSIS PROCEDURE OF BS 6841 FOR EVALUATING

FIGURE 30. VALIDATION OF SENSITIVITY OF PROTOTYPE II BY USING GRAVITATION TEST FOR

FIGURE 37. AN EXAMPLE OF THE EFFECT OF FREQUENCY WEIGHTING W D FOR THE

FIGURE 40. THE RELATIVE CONTRIBUTION (%) OF FREQUENCY WEIGHTED SEAT

FIGURE 46. CORRELATION (SPEARMAN R2) BETWEEN FREQUENCY WEIGHTED OVERALL

FIGURE 47. CORRELATION (SPEARMAN R2) BETWEEN FREQUENCY WEIGHTED OVERALL

FIGURE 48. SIX-AXIS SHAKER AT LOUGHBOROUGH UNIVERSITY FOR CONDUCTING THE

FIGURE 56. THE VISUAL INFORMATION PRESENTED TO PARTICIPANTS TO JUDGE

FIGURE 61. AN EXAMPLE OF CONTINUOUS JUDGMENT OF VIBRATION (LOWER FIGURE SHOWS